Stefanos Nikolaidis

Yulun Zhang, Matthew C. Fontaine, Varun Bhatt et al. · cmu

We study the problem of generating arbitrarily large environments to improve the throughput of multi-robot systems. Prior work proposes Quality Diversity (QD) algorithms as an effective method for optimizing the environments of automated warehouses. However, these approaches optimize only relatively small environments, falling short when it comes to replicating real-world warehouse sizes. The challenge arises from the exponential increase in the search space as the environment size increases. Additionally, the previous methods have only been tested with up to 350 robots in simulations, while practical warehouses could host thousands of robots. In this paper, instead of optimizing environments, we propose to optimize Neural Cellular Automata (NCA) environment generators via QD algorithms. We train a collection of NCA generators with QD algorithms in small environments and then generate arbitrarily large environments from the generators at test time. We show that NCA environment generators maintain consistent, regularized patterns regardless of environment size, significantly enhancing the scalability of multi-robot systems in two different domains with up to 2,350 robots. Additionally, we demonstrate that our method scales a single-agent reinforcement learning policy to arbitrarily large environments with similar patterns. We include the source code at \url{https://github.com/lunjohnzhang/warehouse_env_gen_nca_public}.

Bryon Tjanaka, Matthew C. Fontaine, David H. Lee et al. · cmu

Recent years have seen a rise in the popularity of quality diversity (QD) optimization, a branch of optimization that seeks to find a collection of diverse, high-performing solutions to a given problem. To grow further, we believe the QD community faces two challenges: developing a framework to represent the field's growing array of algorithms, and implementing that framework in software that supports a range of researchers and practitioners. To address these challenges, we have developed pyribs, a library built on a highly modular conceptual QD framework. By replacing components in the conceptual framework, and hence in pyribs, users can compose algorithms from across the QD literature; equally important, they can identify unexplored algorithm variations. Furthermore, pyribs makes this framework simple, flexible, and accessible, with a user-friendly API supported by extensive documentation and tutorials. This paper overviews the creation of pyribs, focusing on the conceptual framework that it implements and the design principles that have guided the library's development.

Cheng Qian, Yulun Zhang, Varun Bhatt et al. · cmu

We use the Quality Diversity (QD) algorithm with Neural Cellular Automata (NCA) to automatically evaluate Multi-Agent Path Finding (MAPF) algorithms by generating diverse maps. Previously, researchers typically evaluate MAPF algorithms on a set of specific, human-designed maps at their initial stage of algorithm design. However, such fixed maps may not cover all scenarios, and algorithms may overfit to the small set of maps. To seek further improvements, systematic evaluations on a diverse suite of maps are needed. In this work, we propose Quality-Diversity Multi-Agent Path Finding Performance EvaluatoR (QD-MAPPER), a general framework that takes advantage of the QD algorithm to comprehensively understand the performance of MAPF algorithms by generating maps with patterns, be able to make fair comparisons between two MAPF algorithms, providing further information on the selection between two algorithms and on the design of the algorithms. Empirically, we employ this technique to evaluate and compare the behavior of different types of MAPF algorithms, including search-based, priority-based, rule-based, and learning-based algorithms. Through both single-algorithm experiments and comparisons between algorithms, researchers can identify patterns that each MAPF algorithm excels and detect disparities in runtime or success rates between different algorithms.

Varun Bhatt, Bryon Tjanaka, Matthew C. Fontaine et al.

Recent progress in reinforcement learning (RL) has started producing generally capable agents that can solve a distribution of complex environments. These agents are typically tested on fixed, human-authored environments. On the other hand, quality diversity (QD) optimization has been proven to be an effective component of environment generation algorithms, which can generate collections of high-quality environments that are diverse in the resulting agent behaviors. However, these algorithms require potentially expensive simulations of agents on newly generated environments. We propose Deep Surrogate Assisted Generation of Environments (DSAGE), a sample-efficient QD environment generation algorithm that maintains a deep surrogate model for predicting agent behaviors in new environments. Results in two benchmark domains show that DSAGE significantly outperforms existing QD environment generation algorithms in discovering collections of environments that elicit diverse behaviors of a state-of-the-art RL agent and a planning agent. Our source code and videos are available at https://dsagepaper.github.io/.

Matthew C. Fontaine, Stefanos Nikolaidis

Single-objective optimization algorithms search for the single highest-quality solution with respect to an objective. Quality diversity (QD) optimization algorithms, such as Covariance Matrix Adaptation MAP-Elites (CMA-ME), search for a collection of solutions that are both high-quality with respect to an objective and diverse with respect to specified measure functions. However, CMA-ME suffers from three major limitations highlighted by the QD community: prematurely abandoning the objective in favor of exploration, struggling to explore flat objectives, and having poor performance for low-resolution archives. We propose a new quality diversity algorithm, Covariance Matrix Adaptation MAP-Annealing (CMA-MAE), that addresses all three limitations. We provide theoretical justifications for the new algorithm with respect to each limitation. Our theory informs our experiments, which support the theory and show that CMA-MAE achieves state-of-the-art performance and robustness.

Yilei Zeng, Jiali Duan, Yang Li et al.

Human-centered AI considers human experiences with AI performance. While abundant research has been helping AI achieve superhuman performance either by fully automatic or weak supervision learning, fewer endeavors are experimenting with how AI can tailor to humans' preferred skill level given fine-grained input. In this work, we guide the curriculum reinforcement learning results towards a preferred performance level that is neither too hard nor too easy via learning from the human decision process. To achieve this, we developed a portable, interactive platform that enables the user to interact with agents online via manipulating the task difficulty, observing performance, and providing curriculum feedback. Our system is highly parallelizable, making it possible for a human to train large-scale reinforcement learning applications that require millions of samples without a server. The result demonstrates the effectiveness of an interactive curriculum for reinforcement learning involving human-in-the-loop. It shows reinforcement learning performance can successfully adjust in sync with the human desired difficulty level. We believe this research will open new doors for achieving flow and personalized adaptive difficulties.

Shivin Dass, Karl Pertsch, Hejia Zhang et al.

Large-scale data is an essential component of machine learning as demonstrated in recent advances in natural language processing and computer vision research. However, collecting large-scale robotic data is much more expensive and slower as each operator can control only a single robot at a time. To make this costly data collection process efficient and scalable, we propose Policy Assisted TeleOperation (PATO), a system which automates part of the demonstration collection process using a learned assistive policy. PATO autonomously executes repetitive behaviors in data collection and asks for human input only when it is uncertain about which subtask or behavior to execute. We conduct teleoperation user studies both with a real robot and a simulated robot fleet and demonstrate that our assisted teleoperation system reduces human operators' mental load while improving data collection efficiency. Further, it enables a single operator to control multiple robots in parallel, which is a first step towards scalable robotic data collection. For code and video results, see https://clvrai.com/pato

Bryon Tjanaka, Matthew C. Fontaine, David H. Lee et al.

Pre-training a diverse set of neural network controllers in simulation has enabled robots to adapt online to damage in robot locomotion tasks. However, finding diverse, high-performing controllers requires expensive network training and extensive tuning of a large number of hyperparameters. On the other hand, Covariance Matrix Adaptation MAP-Annealing (CMA-MAE), an evolution strategies (ES)-based quality diversity algorithm, does not have these limitations and has achieved state-of-the-art performance on standard QD benchmarks. However, CMA-MAE cannot scale to modern neural network controllers due to its quadratic complexity. We leverage efficient approximation methods in ES to propose three new CMA-MAE variants that scale to high dimensions. Our experiments show that the variants outperform ES-based baselines in benchmark robotic locomotion tasks, while being comparable with or exceeding state-of-the-art deep reinforcement learning-based quality diversity algorithms.

Ya-Chuan Hsu, Matthew C. Fontaine, Sam Earle et al.

We present a method for generating arrangements of indoor furniture from human-designed furniture layout data. Our method creates arrangements that target specified diversity, such as the total price of all furniture in the room and the number of pieces placed. To generate realistic furniture arrangement, we train a generative adversarial network (GAN) on human-designed layouts. To target specific diversity in the arrangements, we optimize the latent space of the GAN via a quality diversity algorithm to generate a diverse arrangement collection. Experiments show our approach discovers a set of arrangements that are similar to human-designed layouts but varies in price and number of furniture pieces.

Varun Bhatt, Heramb Nemlekar, Matthew C. Fontaine et al.

As human-robot interaction (HRI) systems advance, so does the difficulty of evaluating and understanding the strengths and limitations of these systems in different environments and with different users. To this end, previous methods have algorithmically generated diverse scenarios that reveal system failures in a shared control teleoperation task. However, these methods require directly evaluating generated scenarios by simulating robot policies and human actions. The computational cost of these evaluations limits their applicability in more complex domains. Thus, we propose augmenting scenario generation systems with surrogate models that predict both human and robot behaviors. In the shared control teleoperation domain and a more complex shared workspace collaboration task, we show that surrogate assisted scenario generation efficiently synthesizes diverse datasets of challenging scenarios. We demonstrate that these failures are reproducible in real-world interactions.

Tianle Huang, Nitish Sontakke, K. Niranjan Kumar et al.

Domain randomization (DR), which entails training a policy with randomized dynamics, has proven to be a simple yet effective algorithm for reducing the gap between simulation and the real world. However, DR often requires careful tuning of randomization parameters. Methods like Bayesian Domain Randomization (Bayesian DR) and Active Domain Randomization (Adaptive DR) address this issue by automating parameter range selection using real-world experience. While effective, these algorithms often require long computation time, as a new policy is trained from scratch every iteration. In this work, we propose Adaptive Bayesian Domain Randomization via Strategic Fine-tuning (BayRnTune), which inherits the spirit of BayRn but aims to significantly accelerate the learning processes by fine-tuning from previously learned policy. This idea leads to a critical question: which previous policy should we use as a prior during fine-tuning? We investigated four different fine-tuning strategies and compared them against baseline algorithms in five simulated environments, ranging from simple benchmark tasks to more complex legged robot environments. Our analysis demonstrates that our method yields better rewards in the same amount of timesteps compared to vanilla domain randomization or Bayesian DR.

Aniruddh G. Puranic, Jyotirmoy V. Deshmukh, Stefanos Nikolaidis

In the learning from demonstration (LfD) paradigm, understanding and evaluating the demonstrated behaviors plays a critical role in extracting control policies for robots. Without this knowledge, a robot may infer incorrect reward functions that lead to undesirable or unsafe control policies. Recent work has proposed an LfD framework where a user provides a set of formal task specifications to guide LfD, to address the challenge of reward shaping. However, in this framework, specifications are manually ordered in a performance graph (a partial order that specifies relative importance between the specifications). The main contribution of this paper is an algorithm to learn the performance graph directly from the user-provided demonstrations, and show that the reward functions generated using the learned performance graph generate similar policies to those from manually specified performance graphs. We perform a user study that shows that priorities specified by users on behaviors in a simulated highway driving domain match the automatically inferred performance graph. This establishes that we can accurately evaluate user demonstrations with respect to task specifications without expert criteria.

Sumeet Batra, Bryon Tjanaka, Stefanos Nikolaidis et al.

Quality Diversity (QD) has shown great success in discovering high-performing, diverse policies for robot skill learning. While current benchmarks have led to the development of powerful QD methods, we argue that new paradigms must be developed to facilitate open-ended search and generalizability. In particular, many methods focus on learning diverse agents that each move to a different xy position in MAP-Elites-style bounded archives. Here, we show that such tasks can be accomplished with a single, goal-conditioned policy paired with a classical planner, achieving O(1) space complexity w.r.t. the number of policies and generalization to task variants. We hypothesize that this approach is successful because it extracts task-invariant structural knowledge by modeling a relational graph between adjacent cells in the archive. We motivate this view with emerging evidence from computational neuroscience and explore connections between QD and models of cognitive maps in human and other animal brains. We conclude with a discussion exploring the relationships between QD and cognitive maps, and propose future research directions inspired by cognitive maps towards future generalizable algorithms capable of truly open-ended search.

Aniruddh G. Puranic, Jyotirmoy V. Deshmukh, Stefanos Nikolaidis

Apprenticeship learning crucially depends on effectively learning rewards, and hence control policies from user demonstrations. Of particular difficulty is the setting where the desired task consists of a number of sub-goals with temporal dependencies. The quality of inferred rewards and hence policies are typically limited by the quality of demonstrations, and poor inference of these can lead to undesirable outcomes. In this letter, we show how temporal logic specifications that describe high level task objectives, are encoded in a graph to define a temporal-based metric that reasons about behaviors of demonstrators and the learner agent to improve the quality of inferred rewards and policies. Through experiments on a diverse set of robot manipulator simulations, we show how our framework overcomes the drawbacks of prior literature by drastically improving the number of demonstrations required to learn a control policy.

Stefanos Nikolaidis

The increasing complexity of robots and autonomous agents that interact with people highlights the critical need for approaches that systematically test them before deployment. This review paper presents a general framework for solving this problem, describes the insights that we have gained from working on each component of the framework, and shows how integrating these components leads to the discovery of a diverse range of realistic and challenging scenarios that reveal previously unknown failures in deployed robotic systems interacting with people.

Saeed Hedayatian, Stefanos Nikolaidis

Quality-Diversity (QD) algorithms constitute a branch of optimization that is concerned with discovering a diverse and high-quality set of solutions to an optimization problem. Current QD methods commonly maintain diversity by dividing the behavior space into discrete regions, ensuring that solutions are distributed across different parts of the space. The QD problem is then solved by searching for the best solution in each region. This approach to QD optimization poses challenges in large solution spaces, where storing many solutions is impractical, and in high-dimensional behavior spaces, where discretization becomes ineffective due to the curse of dimensionality. We present an alternative framing of the QD problem, called \emph{Soft QD}, that sidesteps the need for discretizations. We validate this formulation by demonstrating its desirable properties, such as monotonicity, and by relating its limiting behavior to the widely used QD Score metric. Furthermore, we leverage it to derive a novel differentiable QD algorithm, \emph{Soft QD Using Approximated Diversity (SQUAD)}, and demonstrate empirically that it is competitive with current state of the art methods on standard benchmarks while offering better scalability to higher dimensional problems.

Bryon Tjanaka, Henry Chen, Matthew C. Fontaine et al.

Quality diversity (QD) optimization searches for a collection of solutions that optimize an objective while attaining diverse outputs of a user-specified, vector-valued measure function. Contemporary QD algorithms are typically limited to low-dimensional measures because high-dimensional measures are prone to distortion, where many solutions found by the QD algorithm map to similar measures. For example, the state-of-the-art CMA-MAE algorithm guides measure space exploration with a histogram in measure space that records so-called discount values. However, CMA-MAE stagnates in domains with high-dimensional measure spaces because solutions with similar measures fall into the same histogram cell and hence receive the same discount value. To address these limitations, we propose Discount Model Search (DMS), which guides exploration with a model that provides a smooth, continuous representation of discount values. In high-dimensional measure spaces, this model enables DMS to distinguish between solutions with similar measures and thus continue exploration. We show that DMS facilitates new capabilities for QD algorithms by introducing two new domains where the measure space is the high-dimensional space of images, which enables users to specify their desired measures by providing a dataset of images rather than hand-designing the measure function. Results in these domains and on high-dimensional benchmarks show that DMS outperforms CMA-MAE and other existing black-box QD algorithms.

Yulun Zhang, He Jiang, Varun Bhatt et al. · cmu

We study how to use guidance to improve the throughput of lifelong Multi-Agent Path Finding (MAPF). Previous studies have demonstrated that, while incorporating guidance, such as highways, can accelerate MAPF algorithms, this often results in a trade-off with solution quality. In addition, how to generate good guidance automatically remains largely unexplored, with current methods falling short of surpassing manually designed ones. In this work, we introduce the guidance graph as a versatile representation of guidance for lifelong MAPF, framing Guidance Graph Optimization as the task of optimizing its edge weights. We present two GGO algorithms to automatically generate guidance for arbitrary lifelong MAPF algorithms and maps. The first method directly optimizes edge weights, while the second method optimizes an update model capable of generating edge weights. Empirically, we show that (1) our guidance graphs improve the throughput of three representative lifelong MAPF algorithms in eight benchmark maps, and (2) our update model can generate guidance graphs for as large as $93 \times 91$ maps and as many as 3,000 agents. We include the source code at: \url{https://github.com/lunjohnzhang/ggo_public}. All optimized guidance graphs are available online at: \url{https://yulunzhang.net/publication/zhang2024ggo}.

Allen Chang, Matthew C. Fontaine, Serena Booth et al.

Generative models can serve as surrogates for some real data sources by creating synthetic training datasets, but in doing so they may transfer biases to downstream tasks. We focus on protecting quality and diversity when generating synthetic training datasets. We propose quality-diversity generative sampling (QDGS), a framework for sampling data uniformly across a user-defined measure space, despite the data coming from a biased generator. QDGS is a model-agnostic framework that uses prompt guidance to optimize a quality objective across measures of diversity for synthetically generated data, without fine-tuning the generative model. Using balanced synthetic datasets generated by QDGS, we first debias classifiers trained on color-biased shape datasets as a proof-of-concept. By applying QDGS to facial data synthesis, we prompt for desired semantic concepts, such as skin tone and age, to create an intersectional dataset with a combined blend of visual features. Leveraging this balanced data for training classifiers improves fairness while maintaining accuracy on facial recognition benchmarks. Code available at: https://github.com/Cylumn/qd-generative-sampling.

Nathaniel Dennler, Zhonghao Shi, Stefanos Nikolaidis et al.

Assistive robots interact with humans and must adapt to different users' preferences to be effective. An easy and effective technique to learn non-expert users' preferences is through rankings of robot behaviors, for example, robot movement trajectories or gestures. Existing techniques focus on generating trajectories for users to rank that maximize the outcome of the preference learning process. However, the generated trajectories do not appear to reflect the user's preference over repeated interactions. In this work, we design an algorithm to generate trajectories for users to rank that we call Covariance Matrix Adaptation Evolution Strategies with Information Gain (CMA-ES-IG). CMA-ES-IG prioritizes the user's experience of the preference learning process. We show that users find our algorithm more intuitive and easier to use than previous approaches across both physical and social robot tasks. This project's code is hosted at github.com/interaction-lab/CMA-ES-IG

Yulun Zhang, Matthew C. Fontaine, Varun Bhatt et al.

With the rapid progress in Multi-Agent Path Finding (MAPF), researchers have studied how MAPF algorithms can be deployed to coordinate hundreds of robots in large automated warehouses. While most works try to improve the throughput of such warehouses by developing better MAPF algorithms, we focus on improving the throughput by optimizing the warehouse layout. We show that, even with state-of-the-art MAPF algorithms, commonly used human-designed layouts can lead to congestion for warehouses with large numbers of robots and thus have limited scalability. We extend existing automatic scenario generation methods to optimize warehouse layouts. Results show that our optimized warehouse layouts (1) reduce traffic congestion and thus improve throughput, (2) improve the scalability of the automated warehouses by doubling the number of robots in some cases, and (3) are capable of generating layouts with user-specified diversity measures. We include the source code at: https://github.com/lunjohnzhang/warehouse_env_gen_public

Bryon Tjanaka, Matthew C. Fontaine, Julian Togelius et al.

Consider the problem of training robustly capable agents. One approach is to generate a diverse collection of agent polices. Training can then be viewed as a quality diversity (QD) optimization problem, where we search for a collection of performant policies that are diverse with respect to quantified behavior. Recent work shows that differentiable quality diversity (DQD) algorithms greatly accelerate QD optimization when exact gradients are available. However, agent policies typically assume that the environment is not differentiable. To apply DQD algorithms to training agent policies, we must approximate gradients for performance and behavior. We propose two variants of the current state-of-the-art DQD algorithm that compute gradients via approximation methods common in reinforcement learning (RL). We evaluate our approach on four simulated locomotion tasks. One variant achieves results comparable to the current state-of-the-art in combining QD and RL, while the other performs comparably in two locomotion tasks. These results provide insight into the limitations of current DQD algorithms in domains where gradients must be approximated. Source code is available at https://github.com/icaros-usc/dqd-rl

Yulun Zhang, Matthew C. Fontaine, Amy K. Hoover et al.

We study the problem of efficiently generating high-quality and diverse content in games. Previous work on automated deckbuilding in Hearthstone shows that the quality diversity algorithm MAP-Elites can generate a collection of high-performing decks with diverse strategic gameplay. However, MAP-Elites requires a large number of expensive evaluations to discover a diverse collection of decks. We propose assisting MAP-Elites with a deep surrogate model trained online to predict game outcomes with respect to candidate decks. MAP-Elites discovers a diverse dataset to improve the surrogate model accuracy, while the surrogate model helps guide MAP-Elites towards promising new content. In a Hearthstone deckbuilding case study, we show that our approach improves the sample efficiency of MAP-Elites and outperforms a model trained offline with random decks, as well as a linear surrogate model baseline, setting a new state-of-the-art for quality diversity approaches in automated Hearthstone deckbuilding. We include the source code for all the experiments at: https://github.com/icaros-usc/EvoStone2.

Matthew C. Fontaine, Stefanos Nikolaidis

Quality diversity (QD) is a growing branch of stochastic optimization research that studies the problem of generating an archive of solutions that maximize a given objective function but are also diverse with respect to a set of specified measure functions. However, even when these functions are differentiable, QD algorithms treat them as "black boxes", ignoring gradient information. We present the differentiable quality diversity (DQD) problem, a special case of QD, where both the objective and measure functions are first order differentiable. We then present MAP-Elites via a Gradient Arborescence (MEGA), a DQD algorithm that leverages gradient information to efficiently explore the joint range of the objective and measure functions. Results in two QD benchmark domains and in searching the latent space of a StyleGAN show that MEGA significantly outperforms state-of-the-art QD algorithms, highlighting DQD's promise for efficient quality diversity optimization when gradient information is available. Source code is available at https://github.com/icaros-usc/dqd.

Hejia Zhang, Jie Zhong, Stefanos Nikolaidis

People often watch videos on the web to learn how to cook new recipes, assemble furniture or repair a computer. We wish to enable robots with the very same capability. This is challenging; there is a large variation in manipulation actions and some videos even involve multiple persons, who collaborate by sharing and exchanging objects and tools. Furthermore, the learned representations need to be general enough to be transferable to robotic systems. On the other hand, previous work has shown that the space of human manipulation actions has a linguistic, hierarchical structure that relates actions to manipulated objects and tools. Building upon this theory of language for action, we propose a system for understanding and executing demonstrated action sequences from full-length, real-world cooking videos on the web. The system takes as input a new, previously unseen cooking video annotated with object labels and bounding boxes, and outputs a collaborative manipulation action plan for one or more robotic arms. We demonstrate performance of the system in a standardized dataset of 100 YouTube cooking videos, as well as in six full-length Youtube videos that include collaborative actions between two participants. We compare our system with a baseline system that consists of a state-of-the-art action detection baseline and show our system achieves higher action detection accuracy. We additionally propose an open-source platform for executing the learned plans in a simulation environment as well as with an actual robotic arm.

Enyu Zhao, Vedant Raval, Hejia Zhang et al.

Vision-Language Models (VLMs) have revolutionized artificial intelligence and robotics due to their commonsense reasoning capabilities. In robotic manipulation, VLMs are used primarily as high-level planners, but recent work has also studied their lower-level reasoning ability, which refers to making decisions about precise robot movements. However, the community currently lacks a clear and common benchmark that can evaluate how well VLMs can aid low-level reasoning in robotics. Consequently, we propose a novel benchmark, ManipBench, to evaluate the low-level robot manipulation reasoning capabilities of VLMs across various dimensions, including how well they understand object-object interactions and deformable object manipulation. We extensively test 33 representative VLMs across 10 model families on our benchmark, including variants to test different model sizes. Our evaluation shows that the performance of VLMs significantly varies across tasks, and there is a strong correlation between this performance and trends in our real-world manipulation tasks. It also shows that there remains a significant gap between these models and human-level understanding. See our website at: https://manipbench.github.io.

David H. Lee, Anishalakshmi V. Palaparthi, Matthew C. Fontaine et al.

Diversity optimization seeks to discover a set of solutions that elicit diverse features. Prior work has proposed Novelty Search (NS), which, given a current set of solutions, seeks to expand the set by finding points in areas of low density in the feature space. However, to estimate density, NS relies on a heuristic that considers the k-nearest neighbors of the search point in the feature space, which yields a weaker stability guarantee. We propose Density Descent Search (DDS), an algorithm that explores the feature space via CMA-ES on a continuous density estimate of the feature space that also provides a stronger stability guarantee. We experiment with DDS and two density estimation methods: kernel density estimation (KDE) and continuous normalizing flow (CNF). On several standard diversity optimization benchmarks, DDS outperforms NS, the recently proposed MAP-Annealing algorithm, and other state-of-the-art baselines. Additionally, we prove that DDS with KDE provides stronger stability guarantees than NS, making it more suitable for adaptive optimizers. Furthermore, we prove that NS is a special case of DDS that descends a KDE of the feature space.

Robby Costales, Stefanos Nikolaidis

The wider application of end-to-end learning methods to embodied decision-making domains remains bottlenecked by their reliance on a superabundance of training data representative of the target domain. Meta-reinforcement learning (meta-RL) approaches abandon the aim of zero-shot generalization--the goal of standard reinforcement learning (RL)--in favor of few-shot adaptation, and thus hold promise for bridging larger generalization gaps. While learning this meta-level adaptive behavior still requires substantial data, efficient environment simulators approaching real-world complexity are growing in prevalence. Even so, hand-designing sufficiently diverse and numerous simulated training tasks for these complex domains is prohibitively labor-intensive. Domain randomization (DR) and procedural generation (PG), offered as solutions to this problem, require simulators to possess carefully-defined parameters which directly translate to meaningful task diversity--a similarly prohibitive assumption. In this work, we present DIVA, an evolutionary approach for generating diverse training tasks in such complex, open-ended simulators. Like unsupervised environment design (UED) methods, DIVA can be applied to arbitrary parameterizations, but can additionally incorporate realistically-available domain knowledge--thus inheriting the flexibility and generality of UED, and the supervised structure embedded in well-designed simulators exploited by DR and PG. Our empirical results showcase DIVA's unique ability to overcome complex parameterizations and successfully train adaptive agent behavior, far outperforming competitive baselines from prior literature. These findings highlight the potential of such semi-supervised environment design (SSED) approaches, of which DIVA is the first humble constituent, to enable training in realistic simulated domains, and produce more robust and capable adaptive agents.

Nathaniel Dennler, Stefanos Nikolaidis, Maja Matarić

Adaptive training programs are crucial for recovery post stroke. However, developing programs that automatically adapt depends on quantifying how difficult a task is for a specific individual at a particular stage of their recovery. In this work, we propose a method that automatically generates regions of different task difficulty levels based on an individual's performance. We show that this technique explains the variance in user performance for a reaching task better than previous approaches to estimating task difficulty.

Nathaniel Dennler, Stefanos Nikolaidis, Maja Matarić

Users develop mental models of robots to conceptualize what kind of interactions they can have with those robots. The conceptualizations are often formed before interactions with the robot and are based only on observing the robot's physical design. As a result, understanding conceptualizations formed from physical design is necessary to understand how users intend to interact with the robot. We propose to use multimodal features of robot embodiments to predict what kinds of expectations users will have about a given robot's social and physical capabilities. We show that using such features provides information about general mental models of the robots that generalize across socially interactive robots. We describe how these models can be incorporated into interaction design and physical design for researchers working with socially interactive robots.

Saeed Hedayatian, Stefanos Nikolaidis

Quality-Diversity (QD) algorithms have shown remarkable success in discovering diverse, high-performing solutions, but rely heavily on hand-crafted behavioral descriptors that constrain exploration to predefined notions of diversity. Leveraging the equivalence between policies and occupancy measures, we present a theoretically grounded approach to automatically generate behavioral descriptors by embedding the occupancy measures of policies in Markov Decision Processes. Our method, AutoQD, leverages random Fourier features to approximate the Maximum Mean Discrepancy (MMD) between policy occupancy measures, creating embeddings whose distances reflect meaningful behavioral differences. A low-dimensional projection of these embeddings that captures the most behaviorally significant dimensions is then used as behavioral descriptors for off-the-shelf QD methods. We prove that our embeddings converge to true MMD distances between occupancy measures as the number of sampled trajectories and embedding dimensions increase. Through experiments in multiple continuous control tasks we demonstrate AutoQD's ability in discovering diverse policies without predefined behavioral descriptors, presenting a well-motivated alternative to prior methods in unsupervised Reinforcement Learning and QD optimization. Our approach opens new possibilities for open-ended learning and automated behavior discovery in sequential decision making settings without requiring domain-specific knowledge.

Nathaniel Dennler, Zhonghao Shi, Uksang Yoo et al.

Rehabilitation robots are often used in game-like interactions for rehabilitation to increase a person's motivation to complete rehabilitation exercises. By adjusting exercise difficulty for a specific user throughout the exercise interaction, robots can maximize both the user's rehabilitation outcomes and the their motivation throughout the exercise. Previous approaches have assumed exercises have generic difficulty values that apply to all users equally, however, we identified that stroke survivors have varied and unique perceptions of exercise difficulty. For example, some stroke survivors found reaching vertically more difficult than reaching farther but lower while others found reaching farther more challenging than reaching vertically. In this paper, we formulate a causal tree-based method to calculate exercise difficulty based on the user's performance. We find that this approach accurately models exercise difficulty and provides a readily interpretable model of why that exercise is difficult for both users and caretakers.

Siddharth Srikanth, Varun Bhatt, Boshen Zhang et al.

Understanding how humans collaborate and communicate in teams is essential for improving human-agent teaming and AI-assisted decision-making. However, relying solely on data from large-scale user studies is impractical due to logistical, ethical, and practical constraints, necessitating synthetic models of multiple diverse human behaviors. Recently, agents powered by Large Language Models (LLMs) have been shown to emulate human-like behavior in social settings. But, obtaining a large set of diverse behaviors requires manual effort in the form of designing prompts. On the other hand, Quality Diversity (QD) optimization has been shown to be capable of generating diverse Reinforcement Learning (RL) agent behavior. In this work, we combine QD optimization with LLM-powered agents to iteratively search for prompts that generate diverse team behavior in a long-horizon, multi-step collaborative environment. We first show, through a human-subjects experiment (n=54 participants), that humans exhibit diverse coordination and communication behavior in this domain. We then show that our approach can effectively replicate trends from human teaming data and also capture behaviors that are not easily observed without collecting large amounts of data. Our findings highlight the combination of QD and LLM-powered agents as an effective tool for studying teaming and communication strategies in multi-agent collaboration.

Nathaniel Dennler, Stefanos Nikolaidis, Maja Matarić

People have a variety of preferences for how robots behave. To understand and reason about these preferences, robots aim to learn a reward function that describes how aligned robot behaviors are with a user's preferences. Good representations of a robot's behavior can significantly reduce the time and effort required for a user to teach the robot their preferences. Specifying these representations -- what "features" of the robot's behavior matter to users -- remains a difficult problem; Features learned from raw data lack semantic meaning and features learned from user data require users to engage in tedious labeling processes. Our key insight is that users tasked with customizing a robot are intrinsically motivated to produce labels through exploratory search; they explore behaviors that they find interesting and ignore behaviors that are irrelevant. To harness this novel data source of exploratory actions, we propose contrastive learning from exploratory actions (CLEA) to learn trajectory features that are aligned with features that users care about. We learned CLEA features from exploratory actions users performed in an open-ended signal design activity (N=25) with a Kuri robot, and evaluated CLEA features through a second user study with a different set of users (N=42). CLEA features outperformed self-supervised features when eliciting user preferences over four metrics: completeness, simplicity, minimality, and explainability.

Sumeet Batra, Bryon Tjanaka, Matthew C. Fontaine et al.

Training generally capable agents that thoroughly explore their environment and learn new and diverse skills is a long-term goal of robot learning. Quality Diversity Reinforcement Learning (QD-RL) is an emerging research area that blends the best aspects of both fields -- Quality Diversity (QD) provides a principled form of exploration and produces collections of behaviorally diverse agents, while Reinforcement Learning (RL) provides a powerful performance improvement operator enabling generalization across tasks and dynamic environments. Existing QD-RL approaches have been constrained to sample efficient, deterministic off-policy RL algorithms and/or evolution strategies, and struggle with highly stochastic environments. In this work, we, for the first time, adapt on-policy RL, specifically Proximal Policy Optimization (PPO), to the Differentiable Quality Diversity (DQD) framework and propose additional improvements over prior work that enable efficient optimization and discovery of novel skills on challenging locomotion tasks. Our new algorithm, Proximal Policy Gradient Arborescence (PPGA), achieves state-of-the-art results, including a 4x improvement in best reward over baselines on the challenging humanoid domain.

Nathaniel Dennler, Changxiao Ruan, Jessica Hadiwijoyo et al.

The physical design of a robot suggests expectations of that robot's functionality for human users and collaborators. When those expectations align with the true capabilities of the robot, interaction with the robot is enhanced. However, misalignment of those expectations can result in an unsatisfying interaction. This paper uses Mechanical Turk to evaluate user expectation through the use of design metaphors as applied to a wide range of robot embodiments. The first study (N=382) associates crowd-sourced design metaphors to different robot embodiments. The second study (N=803) assesses initial social expectations of robot embodiments. The final study (N=805) addresses the degree of abstraction of the design metaphors and the functional expectations projected on robot embodiments. Together, these results can guide robot designers toward aligning user expectations with true robot capabilities, facilitating positive human-robot interaction.

Heramb Nemlekar, Runyu Guan, Guanyang Luo et al.

To assist human users according to their individual preference in assembly tasks, robots typically require user demonstrations in the given task. However, providing demonstrations in actual assembly tasks can be tedious and time-consuming. Our thesis is that we can learn user preferences in assembly tasks from demonstrations in a representative canonical task. Inspired by previous work in economy of human movement, we propose to represent user preferences as a linear function of abstract task-agnostic features, such as movement and physical and mental effort required by the user. For each user, we learn their preference from demonstrations in a canonical task and use the learned preference to anticipate their actions in the actual assembly task without any user demonstrations in the actual task. We evaluate our proposed method in a model-airplane assembly study and show that preferences can be effectively transferred from canonical to actual assembly tasks, enabling robots to anticipate user actions.

Sam Earle, Justin Snider, Matthew C. Fontaine et al.

We present a method of generating diverse collections of neural cellular automata (NCA) to design video game levels. While NCAs have so far only been trained via supervised learning, we present a quality diversity (QD) approach to generating a collection of NCA level generators. By framing the problem as a QD problem, our approach can train diverse level generators, whose output levels vary based on aesthetic or functional criteria. To efficiently generate NCAs, we train generators via Covariance Matrix Adaptation MAP-Elites (CMA-ME), a quality diversity algorithm which specializes in continuous search spaces. We apply our new method to generate level generators for several 2D tile-based games: a maze game, Sokoban, and Zelda. Our results show that CMA-ME can generate small NCAs that are diverse yet capable, often satisfying complex solvability criteria for deterministic agents. We compare against a Compositional Pattern-Producing Network (CPPN) baseline trained to produce diverse collections of generators and show that the NCA representation yields a better exploration of level-space.

Heramb Nemlekar, Ziang Liu, Suraj Kothawade et al.

The problem of robotic lime picking is challenging; lime plants have dense foliage which makes it difficult for a robotic arm to grasp a lime without coming in contact with leaves. Existing approaches either do not consider leaves, or treat them as obstacles and completely avoid them, often resulting in undesirable or infeasible plans. We focus on reaching a lime in the presence of dense foliage by considering the leaves of a plant as 'permeable obstacles' with a collision cost. We then adapt the rapidly exploring random tree star (RRT*) algorithm for the problem of fruit harvesting by incorporating the cost of collision with leaves into the path cost. To reduce the time required for finding low-cost paths to goal, we bias the growth of the tree using an artificial potential field (APF). We compare our proposed method with prior work in a 2-D environment and a 6-DOF robot simulation. Our experiments and a real-world demonstration on a robotic lime picking task demonstrate the applicability of our approach.

Nathaniel Dennler, Eura Shin, Maja Matarić et al.

This work introduces an approach for automatic hair combing by a lightweight robot. For people living with limited mobility, dexterity, or chronic fatigue, combing hair is often a difficult task that negatively impacts personal routines. We propose a modular system for enabling general robot manipulators to assist with a hair-combing task. The system consists of three main components. The first component is the segmentation module, which segments the location of hair in space. The second component is the path planning module that proposes automatically-generated paths through hair based on user input. The final component creates a trajectory for the robot to execute. We quantitatively evaluate the effectiveness of the paths planned by the system with 48 users and qualitatively evaluate the system with 30 users watching videos of the robot performing a hair-combing task in the physical world. The system is shown to effectively comb different hairstyles.

Nathaniel Dennler, Catherine Yunis, Jonathan Realmuto et al.

Children and adults with cerebral palsy (CP) can have involuntary upper limb movements as a consequence of the symptoms that characterize their motor disability, leading to difficulties in communicating with caretakers and peers. We describe how a socially assistive robot may help individuals with CP to practice non-verbal communicative gestures using an active orthosis in a one-on-one number-guessing game. We performed a user study and data collection with participants with CP; we found that participants preferred an embodied robot over a screen-based agent, and we used the participant data to train personalized models of participant engagement dynamics that can be used to select personalized robot actions. Our work highlights the benefit of personalized models in the engagement of users with CP with a socially assistive robot and offers design insights for future work in this area.

Matthew C. Fontaine, Ya-Chuan Hsu, Yulun Zhang et al.

When studying robots collaborating with humans, much of the focus has been on robot policies that coordinate fluently with human teammates in collaborative tasks. However, less emphasis has been placed on the effect of the environment on coordination behaviors. To thoroughly explore environments that result in diverse behaviors, we propose a framework for procedural generation of environments that are (1) stylistically similar to human-authored environments, (2) guaranteed to be solvable by the human-robot team, and (3) diverse with respect to coordination measures. We analyze the procedurally generated environments in the Overcooked benchmark domain via simulation and an online user study. Results show that the environments result in qualitatively different emerging behaviors and statistically significant differences in collaborative fluency metrics, even when the robot runs the same planning algorithm.

Heramb Nemlekar, Jignesh Modi, Satyandra K. Gupta et al.

To effectively assist human workers in assembly tasks a robot must proactively offer support by inferring their preferences in sequencing the task actions. Previous work has focused on learning the dominant preferences of human workers for simple tasks largely based on their intended goal. However, people may have preferences at different resolutions: they may share the same high-level preference for the order of the sub-tasks but differ in the sequence of individual actions. We propose a two-stage approach for learning and inferring the preferences of human operators based on the sequence of sub-tasks and actions. We conduct an IKEA assembly study and demonstrate how our approach is able to learn the dominant preferences in a complex task. We show that our approach improves the prediction of human actions through cross-validation. Lastly, we show that our two-stage approach improves the efficiency of task execution in an online experiment, and demonstrate its applicability in a real-world robot-assisted IKEA assembly.

Aniruddh G. Puranic, Jyotirmoy V. Deshmukh, Stefanos Nikolaidis

Learning-from-demonstrations is an emerging paradigm to obtain effective robot control policies for complex tasks via reinforcement learning without the need to explicitly design reward functions. However, it is susceptible to imperfections in demonstrations and also raises concerns of safety and interpretability in the learned control policies. To address these issues, we use Signal Temporal Logic to evaluate and rank the quality of demonstrations. Temporal logic-based specifications allow us to create non-Markovian rewards, and also define interesting causal dependencies between tasks such as sequential task specifications. We validate our approach through experiments on discrete-world and OpenAI Gym environments, and show that our approach outperforms the state-of-the-art Maximum Causal Entropy Inverse Reinforcement Learning.

Matthew Fontaine, Stefanos Nikolaidis

The growth of scale and complexity of interactions between humans and robots highlights the need for new computational methods to automatically evaluate novel algorithms and applications. Exploring diverse scenarios of humans and robots interacting in simulation can improve understanding of the robotic system and avoid potentially costly failures in real-world settings. We formulate this problem as a quality diversity (QD) problem, where the goal is to discover diverse failure scenarios by simultaneously exploring both environments and human actions. We focus on the shared autonomy domain, where the robot attempts to infer the goal of a human operator, and adopt the QD algorithm MAP-Elites to generate scenarios for two published algorithms in this domain: shared autonomy via hindsight optimization and linear policy blending. Some of the generated scenarios confirm previous theoretical findings, while others are surprising and bring about a new understanding of state-of-the-art implementations. Our experiments show that MAP-Elites outperforms Monte-Carlo simulation and optimization based methods in effectively searching the scenario space, highlighting its promise for automatic evaluation of algorithms in human-robot interaction.

Hejia Zhang, Matthew C. Fontaine, Amy K. Hoover et al.

Recent advancements in procedural content generation via machine learning enable the generation of video-game levels that are aesthetically similar to human-authored examples. However, the generated levels are often unplayable without additional editing. We propose a generate-then-repair framework for automatic generation of playable levels adhering to specific styles. The framework constructs levels using a generative adversarial network (GAN) trained with human-authored examples and repairs them using a mixed-integer linear program (MIP) with playability constraints. A key component of the framework is computing minimum cost edits between the GAN generated level and the solution of the MIP solver, which we cast as a minimum cost network flow problem. Results show that the proposed framework generates a diverse range of playable levels, that capture the spatial relationships between objects exhibited in the human-authored levels.

Matthew C. Fontaine, Ruilin Liu, Ahmed Khalifa et al.

Generative adversarial networks (GANs) are quickly becoming a ubiquitous approach to procedurally generating video game levels. While GAN generated levels are stylistically similar to human-authored examples, human designers often want to explore the generative design space of GANs to extract interesting levels. However, human designers find latent vectors opaque and would rather explore along dimensions the designer specifies, such as number of enemies or obstacles. We propose using state-of-the-art quality diversity algorithms designed to optimize continuous spaces, i.e. MAP-Elites with a directional variation operator and Covariance Matrix Adaptation MAP-Elites, to efficiently explore the latent space of a GAN to extract levels that vary across a set of specified gameplay measures. In the benchmark domain of Super Mario Bros, we demonstrate how designers may specify gameplay measures to our system and extract high-quality (playable) levels with a diverse range of level mechanics, while still maintaining stylistic similarity to human authored examples. An online user study shows how the different mechanics of the automatically generated levels affect subjective ratings of their perceived difficulty and appearance.

Yifang Chen, Alex Cuellar, Haipeng Luo et al.

When an AI system interacts with multiple users, it frequently needs to make allocation decisions. For instance, a virtual agent decides whom to pay attention to in a group setting, or a factory robot selects a worker to deliver a part. Demonstrating fairness in decision making is essential for such systems to be broadly accepted. We introduce a Multi-Armed Bandit algorithm with fairness constraints, where fairness is defined as a minimum rate that a task or a resource is assigned to a user. The proposed algorithm uses contextual information about the users and the task and makes no assumptions on how the losses capturing the performance of different users are generated. We provide theoretical guarantees of performance and empirical results from simulation and an online user study. The results highlight the benefit of accounting for contexts in fair decision making, especially when users perform better at some contexts and worse at others.

Matthew C. Fontaine, Julian Togelius, Stefanos Nikolaidis et al.

We focus on the challenge of finding a diverse collection of quality solutions on complex continuous domains. While quality diver-sity (QD) algorithms like Novelty Search with Local Competition (NSLC) and MAP-Elites are designed to generate a diverse range of solutions, these algorithms require a large number of evaluations for exploration of continuous spaces. Meanwhile, variants of the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) are among the best-performing derivative-free optimizers in single-objective continuous domains. This paper proposes a new QD algorithm called Covariance Matrix Adaptation MAP-Elites (CMA-ME). Our new algorithm combines the self-adaptation techniques of CMA-ES with archiving and mapping techniques for maintaining diversity in QD. Results from experiments based on standard continuous optimization benchmarks show that CMA-ME finds better-quality solutions than MAP-Elites; similarly, results on the strategic game Hearthstone show that CMA-ME finds both a higher overall quality and broader diversity of strategies than both CMA-ES and MAP-Elites. Overall, CMA-ME more than doubles the performance of MAP-Elites using standard QD performance metrics. These results suggest that QD algorithms augmented by operators from state-of-the-art optimization algorithms can yield high-performing methods for simultaneously exploring and optimizing continuous search spaces, with significant applications to design, testing, and reinforcement learning among other domains.

Houston Claure, Yifang Chen, Jignesh Modi et al.

How should a robot that collaborates with multiple people decide upon the distribution of resources (e.g. social attention, or parts needed for an assembly)? People are uniquely attuned to how resources are distributed. A decision to distribute more resources to one team member than another might be perceived as unfair with potentially detrimental effects for trust. We introduce a multi-armed bandit algorithm with fairness constraints, where a robot distributes resources to human teammates of different skill levels. In this problem, the robot does not know the skill level of each human teammate, but learns it by observing their performance over time. We define fairness as a constraint on the minimum rate that each human teammate is selected throughout the task. We provide theoretical guarantees on performance and perform a large-scale user study, where we adjust the level of fairness in our algorithm. Results show that fairness in resource distribution has a significant effect on users' trust in the system.