Diego Romeres

Tim Schneider, Boris Belousov, Georgia Chalvatzaki et al.

Robotic manipulation stands as a largely unsolved problem despite significant advances in robotics and machine learning in recent years. One of the key challenges in manipulation is the exploration of the dynamics of the environment when there is continuous contact between the objects being manipulated. This paper proposes a model-based active exploration approach that enables efficient learning in sparse-reward robotic manipulation tasks. The proposed method estimates an information gain objective using an ensemble of probabilistic models and deploys model predictive control (MPC) to plan actions online that maximize the expected reward while also performing directed exploration. We evaluate our proposed algorithm in simulation and on a real robot, trained from scratch with our method, on a challenging ball pushing task on tilted tables, where the target ball position is not known to the agent a-priori. Our real-world robot experiment serves as a fundamental application of active exploration in model-based reinforcement learning of complex robotic manipulation tasks.

Seiji Shaw, Devesh K. Jha, Arvind Raghunathan et al.

Dynamic movement primitives are widely used for learning skills which can be demonstrated to a robot by a skilled human or controller. While their generalization capabilities and simple formulation make them very appealing to use, they possess no strong guarantees to satisfy operational safety constraints for a task. In this paper, we present constrained dynamic movement primitives (CDMP) which can allow for constraint satisfaction in the robot workspace. We present a formulation of a non-linear optimization to perturb the DMP forcing weights regressed by locally-weighted regression to admit a Zeroing Barrier Function (ZBF), which certifies workspace constraint satisfaction. We demonstrate the proposed CDMP under different constraints on the end-effector movement such as obstacle avoidance and workspace constraints on a physical robot. A video showing the implementation of the proposed algorithm using different manipulators in different environments could be found here https://youtu.be/hJegJJkJfys.

Yuki Shirai, Devesh K. Jha, Arvind Raghunathan et al.

Generalizable manipulation requires that robots be able to interact with novel objects and environment. This requirement makes manipulation extremely challenging as a robot has to reason about complex frictional interaction with uncertainty in physical properties of the object. In this paper, we study robust optimization for control of pivoting manipulation in the presence of uncertainties. We present insights about how friction can be exploited to compensate for the inaccuracies in the estimates of the physical properties during manipulation. In particular, we derive analytical expressions for stability margin provided by friction during pivoting manipulation. This margin is then used in a bilevel trajectory optimization algorithm to design a controller that maximizes this stability margin to provide robustness against uncertainty in physical properties of the object. We demonstrate our proposed method using a 6 DoF manipulator for manipulating several different objects.

Giulio Giacomuzzos, Ruggero Carli, Diego Romeres et al.

Learning the inverse dynamics of robots directly from data, adopting a black-box approach, is interesting for several real-world scenarios where limited knowledge about the system is available. In this paper, we propose a black-box model based on Gaussian Process (GP) Regression for the identification of the inverse dynamics of robotic manipulators. The proposed model relies on a novel multidimensional kernel, called \textit{Lagrangian Inspired Polynomial} (\kernelInitials{}) kernel. The \kernelInitials{} kernel is based on two main ideas. First, instead of directly modeling the inverse dynamics components, we model as GPs the kinetic and potential energy of the system. The GP prior on the inverse dynamics components is derived from those on the energies by applying the properties of GPs under linear operators. Second, as regards the energy prior definition, we prove a polynomial structure of the kinetic and potential energy, and we derive a polynomial kernel that encodes this property. As a consequence, the proposed model allows also to estimate the kinetic and potential energy without requiring any label on these quantities. Results on simulation and on two real robotic manipulators, namely a 7 DOF Franka Emika Panda, and a 6 DOF MELFA RV4FL, show that the proposed model outperforms state-of-the-art black-box estimators based both on Gaussian Processes and Neural Networks in terms of accuracy, generality and data efficiency. The experiments on the MELFA robot also demonstrate that our approach achieves performance comparable to fine-tuned model-based estimators, despite requiring less prior information.

Chiori Hori, Puyuan Peng, David Harwath et al.

To realize human-robot collaboration, robots need to execute actions for new tasks according to human instructions given finite prior knowledge. Human experts can share their knowledge of how to perform a task with a robot through multi-modal instructions in their demonstrations, showing a sequence of short-horizon steps to achieve a long-horizon goal. This paper introduces a method for robot action sequence generation from instruction videos using (1) an audio-visual Transformer that converts audio-visual features and instruction speech to a sequence of robot actions called dynamic movement primitives (DMPs) and (2) style-transfer-based training that employs multi-task learning with video captioning and weakly-supervised learning with a semantic classifier to exploit unpaired video-action data. We built a system that accomplishes various cooking actions, where an arm robot executes a DMP sequence acquired from a cooking video using the audio-visual Transformer. Experiments with Epic-Kitchen-100, YouCookII, QuerYD, and in-house instruction video datasets show that the proposed method improves the quality of DMP sequences by 2.3 times the METEOR score obtained with a baseline video-to-action Transformer. The model achieved 32% of the task success rate with the task knowledge of the object.

Yuki Shirai, Devesh K. Jha, Arvind Raghunathan et al.

This paper presents a chance-constrained formulation for robust trajectory optimization during manipulation. In particular, we present a chance-constrained optimization for Stochastic Discrete-time Linear Complementarity Systems (SDLCS). To solve the optimization problem, we formulate Mixed-Integer Quadratic Programming with Chance Constraints (MIQPCC). In our formulation, we explicitly consider joint chance constraints for complementarity as well as states to capture the stochastic evolution of dynamics. We evaluate robustness of our optimized trajectories in simulation on several systems. The proposed approach outperforms some recent approaches for robust trajectory optimization for SDLCS.

Fabio Amadio, Alberto Dalla Libera, Daniel Nikovski et al.

We propose a Model-Based Reinforcement Learning (MBRL) algorithm named VF-MC-PILCO, specifically designed for application to mechanical systems where velocities cannot be directly measured. This circumstance, if not adequately considered, can compromise the success of MBRL approaches. To cope with this problem, we define a velocity-free state formulation which consists of the collection of past positions and inputs. Then, VF-MC-PILCO uses Gaussian Process Regression to model the dynamics of the velocity-free state and optimizes the control policy through a particle-based policy gradient approach. We compare VF-MC-PILCO with our previous MBRL algorithm, MC-PILCO4PMS, which handles the lack of direct velocity measurements by modeling the presence of velocity estimators. Results on both simulated (cart-pole and UR5 robot) and real mechanical systems (Furuta pendulum and a ball-and-plate rig) show that the two algorithms achieve similar results. Conveniently, VF-MC-PILCO does not require the design and implementation of state estimators, which can be a challenging and time-consuming activity to be performed by an expert user.

Devesh K. Jha, Siddarth Jain, Diego Romeres et al.

Robots have been steadily increasing their presence in our daily lives, where they can work along with humans to provide assistance in various tasks on industry floors, in offices, and in homes. Automated assembly is one of the key applications of robots, and the next generation assembly systems could become much more efficient by creating collaborative human-robot systems. However, although collaborative robots have been around for decades, their application in truly collaborative systems has been limited. This is because a truly collaborative human-robot system needs to adjust its operation with respect to the uncertainty and imprecision in human actions, ensure safety during interaction, etc. In this paper, we present a system for human-robot collaborative assembly using learning from demonstration and pose estimation, so that the robot can adapt to the uncertainty caused by the operation of humans. Learning from demonstration is used to generate motion trajectories for the robot based on the pose estimate of different goal locations from a deep learning-based vision system. The proposed system is demonstrated using a physical 6 DoF manipulator in a collaborative human-robot assembly scenario. We show successful generalization of the system's operation to changes in the initial and final goal locations through various experiments.

Zixing Wang, Devesh K. Jha, Ahmed H. Qureshi et al.

Diffusion policies have shown to be very efficient at learning complex, multi-modal behaviors for robotic manipulation. However, errors in generated action sequences can compound over time which can potentially lead to failure. Some approaches mitigate this by augmenting datasets with expert demonstrations or learning predictive world models which might be computationally expensive. We introduce Performance Predictive Guidance (PPGuide), a lightweight, classifier-based framework that steers a pre-trained diffusion policy away from failure modes at inference time. PPGuide makes use of a novel self-supervised process: it uses attention-based multiple instance learning to automatically estimate which observation-action chunks from the policy's rollouts are relevant to success or failure. We then train a performance predictor on this self-labeled data. During inference, this predictor provides a real-time gradient to guide the policy toward more robust actions. We validated our proposed PPGuide across a diverse set of tasks from the Robomimic and MimicGen benchmarks, demonstrating consistent improvements in performance.

Kei Suzuki, Jing Liu, Ye Wang et al.

Cross-robot policy learning -- training a single policy to perform well across multiple embodiments -- remains a central challenge in robot learning. Transformer-based policies, such as vision-language-action (VLA) models, are typically embodiment-agnostic and must infer kinematic structure purely from observations, which can reduce robustness across embodiments and even limit performance within a single embodiment. We propose an embodiment-aware transformer policy that injects morphology via three mechanisms: (1) kinematic tokens that factorize actions across joints and compress time through per-joint temporal chunking; (2) a topology-aware attention bias that encodes kinematic topology as an inductive bias in self-attention, encouraging message passing along kinematic edges; and (3) joint-attribute conditioning that augments topology with per-joint descriptors to capture semantics beyond connectivity. Across a range of embodiments, this structured integration consistently improves performance over a vanilla pi0.5 VLA baseline, indicating improved robustness both within an embodiment and across embodiments.

Arvind U. Raghunathan, Devesh K. Jha, Diego Romeres

PYROBOCOP is a lightweight Python-based package for control and optimization of robotic systems described by nonlinear Differential Algebraic Equations (DAEs). In particular, the package can handle systems with contacts that are described by complementarity constraints and provides a general framework for specifying obstacle avoidance constraints. The package performs direct transcription of the DAEs into a set of nonlinear equations by performing orthogonal collocation on finite elements. The resulting optimization problem belongs to the class of Mathematical Programs with Complementarity Constraints (MPCCs). MPCCs fail to satisfy commonly assumed constraint qualifications and require special handling of the complementarity constraints in order for NonLinear Program (NLP) solvers to solve them effectively. PYROBOCOP provides automatic reformulation of the complementarity constraints that enables NLP solvers to perform optimization of robotic systems. The package is interfaced with ADOLC for obtaining sparse derivatives by automatic differentiation and IPOPT for performing optimization. We demonstrate the effectiveness of our approach in terms of speed and flexibility. We provide several numerical examples for several robotic systems with collision avoidance as well as contact constraints represented using complementarity constraints. We provide comparisons with other open source optimization packages like CasADi and Pyomo .

Lingfeng Sun, Devesh K. Jha, Chiori Hori et al. · berkeley

Designing robotic agents to perform open vocabulary tasks has been the long-standing goal in robotics and AI. Recently, Large Language Models (LLMs) have achieved impressive results in creating robotic agents for performing open vocabulary tasks. However, planning for these tasks in the presence of uncertainties is challenging as it requires \enquote{chain-of-thought} reasoning, aggregating information from the environment, updating state estimates, and generating actions based on the updated state estimates. In this paper, we present an interactive planning technique for partially observable tasks using LLMs. In the proposed method, an LLM is used to collect missing information from the environment using a robot and infer the state of the underlying problem from collected observations while guiding the robot to perform the required actions. We also use a fine-tuned Llama 2 model via self-instruct and compare its performance against a pre-trained LLM like GPT-4. Results are demonstrated on several tasks in simulation as well as real-world environments. A video describing our work along with some results could be found here.

Anoop Cherian, Radu Corcodel, Siddarth Jain et al.

Physical reasoning is an important skill needed for robotic agents when operating in the real world. However, solving such reasoning problems often involves hypothesizing and reflecting over complex multi-body interactions under the effect of a multitude of physical forces and thus learning all such interactions poses a significant hurdle for state-of-the-art machine learning frameworks, including large language models (LLMs). To study this problem, we propose a new physical reasoning task and a dataset, dubbed TraySim. Our task involves predicting the dynamics of several objects on a tray that is given an external impact -- the domino effect of the ensued object interactions and their dynamics thus offering a challenging yet controlled setup, with the goal of reasoning being to infer the stability of the objects after the impact. To solve this complex physical reasoning task, we present LLMPhy, a zero-shot black-box optimization framework that leverages the physics knowledge and program synthesis abilities of LLMs, and synergizes these abilities with the world models built into modern physics engines. Specifically, LLMPhy uses an LLM to generate code to iteratively estimate the physical hyperparameters of the system (friction, damping, layout, etc.) via an implicit analysis-by-synthesis approach using a (non-differentiable) simulator in the loop and uses the inferred parameters to imagine the dynamics of the scene towards solving the reasoning task. To show the effectiveness of LLMPhy, we present experiments on our TraySim dataset to predict the steady-state poses of the objects. Our results show that the combination of the LLM and the physics engine leads to state-of-the-art zero-shot physical reasoning performance, while demonstrating superior convergence against standard black-box optimization methods and better estimation of the physical parameters.

Xinghao Zhu, Devesh K. Jha, Diego Romeres et al. · berkeley

Automating the assembly of objects from their parts is a complex problem with innumerable applications in manufacturing, maintenance, and recycling. Unlike existing research, which is limited to target segmentation, pose regression, or using fixed target blueprints, our work presents a holistic multi-level framework for part assembly planning consisting of part assembly sequence inference, part motion planning, and robot contact optimization. We present the Part Assembly Sequence Transformer (PAST) -- a sequence-to-sequence neural network -- to infer assembly sequences recursively from a target blueprint. We then use a motion planner and optimization to generate part movements and contacts. To train PAST, we introduce D4PAS: a large-scale Dataset for Part Assembly Sequences (D4PAS) consisting of physically valid sequences for industrial objects. Experimental results show that our approach generalizes better than prior methods while needing significantly less computational time for inference.

Yuxin Chen, Devesh K. Jha, Masayoshi Tomizuka et al.

Imitation learning from human demonstrations enables robots to perform complex manipulation tasks and has recently witnessed huge success. However, these techniques often struggle to adapt behavior to new preferences or changes in the environment. To address these limitations, we propose Fine-tuning Diffusion Policy with Human Preference (FDPP). FDPP learns a reward function through preference-based learning. This reward is then used to fine-tune the pre-trained policy with reinforcement learning (RL), resulting in alignment of pre-trained policy with new human preferences while still solving the original task. Our experiments across various robotic tasks and preferences demonstrate that FDPP effectively customizes policy behavior without compromising performance. Additionally, we show that incorporating Kullback-Leibler (KL) regularization during fine-tuning prevents over-fitting and helps maintain the competencies of the initial policy.

Shivam Vats, Devesh K. Jha, Maxim Likhachev et al.

Model-based planners and controllers are commonly used to solve complex manipulation problems as they can efficiently optimize diverse objectives and generalize to long horizon tasks. However, they often fail during deployment due to noisy actuation, partial observability and imperfect models. To enable a robot to recover from such failures, we propose to use hierarchical reinforcement learning to learn a recovery policy. The recovery policy is triggered when a failure is detected based on sensory observations and seeks to take the robot to a state from which it can complete the task using the nominal model-based controllers. Our approach, called RecoveryChaining, uses a hybrid action space, where the model-based controllers are provided as additional \emph{nominal} options which allows the recovery policy to decide how to recover, when to switch to a nominal controller and which controller to switch to even with \emph{sparse rewards}. We evaluate our approach in three multi-step manipulation tasks with sparse rewards, where it learns significantly more robust recovery policies than those learned by baselines. We successfully transfer recovery policies learned in simulation to a physical robot to demonstrate the feasibility of sim-to-real transfer with our method.

Joshua Hang Sai Ip, Ankush Chakrabarty, Ali Mesbah et al.

Incorporating user preferences into multi-objective Bayesian optimization (MOBO) allows for personalization of the optimization procedure. Preferences are often abstracted in the form of an unknown utility function, estimated through pairwise comparisons of potential outcomes. However, utility-driven MOBO methods can yield solutions that are dominated by nearby solutions, as non-dominance is not enforced. Additionally, classical MOBO commonly relies on estimating the entire Pareto-front to identify the Pareto-optimal solutions, which can be expensive and ignore user preferences. Here, we present a new method, termed preference-utility-balanced MOBO (PUB-MOBO), that allows users to disambiguate between near-Pareto candidate solutions. PUB-MOBO combines utility-based MOBO with local multi-gradient descent to refine user-preferred solutions to be near-Pareto-optimal. To this end, we propose a novel preference-dominated utility function that concurrently preserves user-preferences and dominance amongst candidate solutions. A key advantage of PUB-MOBO is that the local search is restricted to a (small) region of the Pareto-front directed by user preferences, alleviating the need to estimate the entire Pareto-front. PUB-MOBO is tested on three synthetic benchmark problems: DTLZ1, DTLZ2 and DH1, as well as on three real-world problems: Vehicle Safety, Conceptual Marine Design, and Car Side Impact. PUB-MOBO consistently outperforms state-of-the-art competitors in terms of proximity to the Pareto-front and utility regret across all the problems.

Chiori Hori, Yoshiki Masuyama, Siddarth Jain et al.

Human-robot collaboration towards a shared goal requires robots to understand human action and interaction with the surrounding environment. This paper focuses on human-robot interaction (HRI) based on human-robot dialogue that relies on the robot action confirmation and action step generation using multimodal scene understanding. The state-of-the-art approach uses multimodal transformers to generate robot action steps aligned with robot action confirmation from a single clip showing a task composed of multiple micro steps. Although actions towards a long-horizon task depend on each other throughout an entire video, the current approaches mainly focus on clip-level processing and do not leverage long-context information. This paper proposes a long-context Q-former incorporating left and right context dependency in full videos. Furthermore, this paper proposes a text-conditioning approach to feed text embeddings directly into the LLM decoder to mitigate the high abstraction of the information in text by Q-former. Experiments with the YouCook2 corpus show that the accuracy of confirmation generation is a major factor in the performance of action planning. Furthermore, we demonstrate that the long-context Q-former improves the confirmation and action planning by integrating VideoLLaMA3.

Yuki Shirai, Kei Ota, Devesh K. Jha et al.

Non-prehensile manipulation is challenging due to complex contact interactions between objects, the environment, and robots. Model-based approaches can efficiently generate complex trajectories of robots and objects under contact constraints. However, they tend to be sensitive to model inaccuracies and require access to privileged information (e.g., object mass, size, pose), making them less suitable for novel objects. In contrast, learning-based approaches are typically more robust to modeling errors but require large amounts of data. In this paper, we bridge these two approaches to propose a framework for learning closed-loop pivoting manipulation. By leveraging computationally efficient Contact-Implicit Trajectory Optimization (CITO), we design demonstration-guided deep Reinforcement Learning (RL), leading to sample-efficient learning. We also present a sim-to-real transfer approach using a privileged training strategy, enabling the robot to perform pivoting manipulation using only proprioception, vision, and force sensing without access to privileged information. Our method is evaluated on several pivoting tasks, demonstrating that it can successfully perform sim-to-real transfer. The overview of our method and the hardware experiments are shown at https://youtu.be/akjGDgfwLbM?si=QVw6ExoPy2VsU2g6

Niccolò Turcato, Marco Calì, Alberto Dalla Libera et al.

This short paper describes our proposed solution for the third edition of the "AI Olympics with RealAIGym" competition, held at ICRA 2025. We employed Monte-Carlo Probabilistic Inference for Learning Control (MC-PILCO), an MBRL algorithm recognized for its exceptional data efficiency across various low-dimensional robotic tasks, including cart-pole, ball \& plate, and Furuta pendulum systems. MC-PILCO optimizes a system dynamics model using interaction data, enabling policy refinement through simulation rather than direct system data optimization. This approach has proven highly effective in physical systems, offering greater data efficiency than Model-Free (MF) alternatives. Notably, MC-PILCO has previously won the first two editions of this competition, demonstrating its robustness in both simulated and real-world environments. Besides briefly reviewing the algorithm, we discuss the most critical aspects of the MC-PILCO implementation in the tasks at hand: learning a global policy for the pendubot and acrobot systems.

Devesh K. Jha, Diego Romeres, William Yerazunis et al.

We present the design of a learning-based compliance controller for assembly operations for industrial robots. We propose a solution within the general setting of learning from demonstration (LfD), where a nominal trajectory is provided through demonstration by an expert teacher. This can be used to learn a suitable representation of the skill that can be generalized to novel positions of one of the parts involved in the assembly, for example the hole in a peg-in-hole (PiH) insertion task. Under the expectation that this novel position might not be entirely accurately estimated by a vision or other sensing system, the robot will need to further modify the generated trajectory in response to force readings measured by means of a force-torque (F/T) sensor mounted at the wrist of the robot or another suitable location. Under the assumption of constant velocity of traversing the reference trajectory during assembly, we propose a novel accommodation force controller that allows the robot to safely explore different contact configurations. The data collected using this controller is used to train a Gaussian process model to predict the misalignment in the position of the peg with respect to the target hole. We show that the proposed learning-based approach can correct various contact configurations caused by misalignment between the assembled parts in a PiH task, achieving high success rate during insertion. We show results using an industrial manipulator arm, and demonstrate that the proposed method can perform adaptive insertion using force feedback from the trained machine learning models.

Shiyu Jin, Diego Romeres, Arvind Ragunathan et al.

This paper presents a novel trajectory optimization formulation to solve the robotic assembly of the belt drive unit. Robotic manipulations involving contacts and deformable objects are challenging in both dynamic modeling and trajectory planning. For modeling, variations in the belt tension and contact forces between the belt and the pulley could dramatically change the system dynamics. For trajectory planning, it is computationally expensive to plan trajectories for such hybrid dynamical systems as it usually requires planning for discrete modes separately. In this work, we formulate the belt drive unit assembly task as a trajectory optimization problem with complementarity constraints to avoid explicitly imposing contact mode sequences. The problem is solved as a mathematical program with complementarity constraints (MPCC) to obtain feasible and efficient assembly trajectories. We validate the proposed method both in simulations with a physics engine and in real-world experiments with a robotic manipulator.

Alberto Dalla Libera, Fabio Amadio, Daniel Nikovski et al.

In this paper, we consider the use of black-box Gaussian process (GP) models for trajectory tracking control based on feedback linearization, in the context of mechanical systems. We considered two strategies. The first computes the control input directly by using the GP model, whereas the second computes the input after estimating the individual components of the dynamics. We tested the two strategies on a simulated manipulator with seven degrees of freedom, also varying the GP kernel choice. Results show that the second implementation is more robust w.r.t. the kernel choice and model inaccuracies. Moreover, as regards the choice of kernel, the obtained performance shows that the use of a structured kernel, such as a polynomial kernel, is advantageous, because of its effectiveness with both strategies.

Siyuan Dong, Devesh K. Jha, Diego Romeres et al.

Object insertion is a classic contact-rich manipulation task. The task remains challenging, especially when considering general objects of unknown geometry, which significantly limits the ability to understand the contact configuration between the object and the environment. We study the problem of aligning the object and environment with a tactile-based feedback insertion policy. The insertion process is modeled as an episodic policy that iterates between insertion attempts followed by pose corrections. We explore different mechanisms to learn such a policy based on Reinforcement Learning. The key contribution of this paper is to demonstrate that it is possible to learn a tactile insertion policy that generalizes across different object geometries, and an ablation study of the key design choices for the learning agent: 1) the type of learning scheme: supervised vs. reinforcement learning; 2) the type of learning schedule: unguided vs. curriculum learning; 3) the type of sensing modality: force/torque (F/T) vs. tactile; and 4) the type of tactile representation: tactile RGB vs. tactile flow. We show that the optimal configuration of the learning agent (RL + curriculum + tactile flow) exposed to 4 training objects yields an insertion policy that inserts 4 novel objects with over 85.0% success rate and within 3~4 attempts. Comparisons between F/T and tactile sensing, shows that while an F/T-based policy learns more efficiently, a tactile-based policy provides better generalization.

Fabio Amadio, Alberto Dalla Libera, Riccardo Antonello et al.

In this paper, we present a Model-Based Reinforcement Learning (MBRL) algorithm named \emph{Monte Carlo Probabilistic Inference for Learning COntrol} (MC-PILCO). The algorithm relies on Gaussian Processes (GPs) to model the system dynamics and on a Monte Carlo approach to estimate the policy gradient. This defines a framework in which we ablate the choice of the following components: (i) the selection of the cost function, (ii) the optimization of policies using dropout, (iii) an improved data efficiency through the use of structured kernels in the GP models. The combination of the aforementioned aspects affects dramatically the performance of MC-PILCO. Numerical comparisons in a simulated cart-pole environment show that MC-PILCO exhibits better data efficiency and control performance w.r.t. state-of-the-art GP-based MBRL algorithms. Finally, we apply MC-PILCO to real systems, considering in particular systems with partially measurable states. We discuss the importance of modeling both the measurement system and the state estimators during policy optimization. The effectiveness of the proposed solutions has been tested in simulation and on two real systems, a Furuta pendulum and a ball-and-plate rig.

Fabio Amadio, Alberto Dalla Libera, Ruggero Carli et al.

In this paper, we propose a Model-Based Reinforcement Learning (MBRL) algorithm for Partially Measurable Systems (PMS), i.e., systems where the state can not be directly measured, but must be estimated through proper state observers. The proposed algorithm, named Monte Carlo Probabilistic Inference for Learning COntrol for Partially Measurable Systems (MC-PILCO4PMS), relies on Gaussian Processes (GPs) to model the system dynamics, and on a Monte Carlo approach to update the policy parameters. W.r.t. previous GP-based MBRL algorithms, MC-PILCO4PMS models explicitly the presence of state observers during policy optimization, allowing to deal PMS. The effectiveness of the proposed algorithm has been tested both in simulation and in two real systems.

Kei Ota, Devesh K. Jha, Diego Romeres et al.

Humans quickly solve tasks in novel systems with complex dynamics, without requiring much interaction. While deep reinforcement learning algorithms have achieved tremendous success in many complex tasks, these algorithms need a large number of samples to learn meaningful policies. In this paper, we present a task for navigating a marble to the center of a circular maze. While this system is very intuitive and easy for humans to solve, it can be very difficult and inefficient for standard reinforcement learning algorithms to learn meaningful policies. We present a model that learns to move a marble in the complex environment within minutes of interacting with the real system. Learning consists of initializing a physics engine with parameters estimated using data from the real system. The error in the physics engine is then corrected using Gaussian process regression, which is used to model the residual between real observations and physics engine simulations. The physics engine augmented with the residual model is then used to control the marble in the maze environment using a model-predictive feedback over a receding horizon. To the best of our knowledge, this is the first time that a hybrid model consisting of a full physics engine along with a statistical function approximator has been used to control a complex physical system in real-time using nonlinear model-predictive control (NMPC).

Yifang Liu, Diego Romeres, Devesh K. Jha et al.

One of the main challenges in peg-in-a-hole (PiH) insertion tasks is in handling the uncertainty in the location of the target hole. In order to address it, high-dimensional sensor inputs from sensor modalities such as vision, force/torque sensing, and proprioception can be combined to learn control policies that are robust to this uncertainty in the target pose. Whereas deep learning has shown success in recognizing objects and making decisions with high-dimensional inputs, the learning procedure might damage the robot when applying directly trial- and-error algorithms on the real system. At the same time, learning from Demonstration (LfD) methods have been shown to achieve compelling performance in real robotic systems by leveraging demonstration data provided by experts. In this paper, we investigate the merits of multiple sensor modalities such as vision, force/torque sensors, and proprioception when combined to learn a controller for real world assembly operation tasks using LfD techniques. The study is limited to PiH insertions; we plan to extend the study to more experiments in the future. Additionally, we propose a multi-step-ahead loss function to improve the performance of the behavioral cloning method. Experimental results on a real manipulator support our findings, and show the effectiveness of the proposed loss function.

Alberto Dalla Libera, Diego Romeres, Devesh K. Jha et al.

In this paper, we propose a derivative-free model learning framework for Reinforcement Learning (RL) algorithms based on Gaussian Process Regression (GPR). In many mechanical systems, only positions can be measured by the sensing instruments. Then, instead of representing the system state as suggested by the physics with a collection of positions, velocities, and accelerations, we define the state as the set of past position measurements. However, the equation of motions derived by physical first principles cannot be directly applied in this framework, being functions of velocities and accelerations. For this reason, we introduce a novel derivative-free physically-inspired kernel, which can be easily combined with nonparametric derivative-free Gaussian Process models. Tests performed on two real platforms show that the considered state definition combined with the proposed model improves estimation performance and data-efficiency w.r.t. traditional models based on GPR. Finally, we validate the proposed framework by solving two RL control problems for two real robotic systems.

Patrik Kolaric, Devesh K. Jha, Arvind U. Raghunathan et al.

The goal of this paper is to present a method for simultaneous trajectory and local stabilizing policy optimization to generate local policies for trajectory-centric model-based reinforcement learning (MBRL). This is motivated by the fact that global policy optimization for non-linear systems could be a very challenging problem both algorithmically and numerically. However, a lot of robotic manipulation tasks are trajectory-centric, and thus do not require a global model or policy. Due to inaccuracies in the learned model estimates, an open-loop trajectory optimization process mostly results in very poor performance when used on the real system. Motivated by these problems, we try to formulate the problem of trajectory optimization and local policy synthesis as a single optimization problem. It is then solved simultaneously as an instance of nonlinear programming. We provide some results for analysis as well as achieved performance of the proposed technique under some simplifying assumptions.

Devesh Jha, Arvind Raghunathan, Diego Romeres

We propose a trust region method for policy optimization that employs Quasi-Newton approximation for the Hessian, called Quasi-Newton Trust Region Policy Optimization QNTRPO. Gradient descent is the de facto algorithm for reinforcement learning tasks with continuous controls. The algorithm has achieved state-of-the-art performance when used in reinforcement learning across a wide range of tasks. However, the algorithm suffers from a number of drawbacks including: lack of stepsize selection criterion, and slow convergence. We investigate the use of a trust region method using dogleg step and a Quasi-Newton approximation for the Hessian for policy optimization. We demonstrate through numerical experiments over a wide range of challenging continuous control tasks that our particular choice is efficient in terms of number of samples and improves performance

Jonathan Chang, Nishanth Kumar, Sean Hastings et al.

Robots need to learn skills that can not only generalize across similar problems but also be directed to a specific goal. Previous methods either train a new skill for every different goal or do not infer the specific target in the presence of multiple goals from visual data. We introduce an end-to-end method that represents targetable visuomotor skills as a goal-parameterized neural network policy. By training on an informative subset of available goals with the associated target parameters, we are able to learn a policy that can zero-shot generalize to previously unseen goals. We evaluate our method in a representative 2D simulation of a button-grid and on both button-pressing and peg-insertion tasks on two different physical arms. We demonstrate that our model trained on 33% of the possible goals is able to generalize to more than 90% of the targets in the scene for both simulation and robot experiments. We also successfully learn a mapping from target pixel coordinates to a robot policy to complete a specified goal.

Diego Romeres, Mattia Zorzi, Raffaello Camoriano et al.

This paper discusses online algorithms for inverse dynamics modelling in robotics. Several model classes including rigid body dynamics (RBD) models, data-driven models and semiparametric models (which are a combination of the previous two classes) are placed in a common framework. While model classes used in the literature typically exploit joint velocities and accelerations, which need to be approximated resorting to numerical differentiation schemes, in this paper a new `derivative-free' framework is proposed that does not require this preprocessing step. An extensive experimental study with real data from the right arm of the iCub robot is presented, comparing different model classes and estimation procedures, showing that the proposed `derivative-free' methods outperform existing methodologies.

Diego Romeres, Devesh Jha, Alberto Dalla Libera et al.

This paper presents a problem of model learning for the purpose of learning how to navigate a ball to a goal state in a circular maze environment with two degrees of freedom. The motion of the ball in the maze environment is influenced by several non-linear effects such as dry friction and contacts, which are difficult to model physically. We propose a semiparametric model to estimate the motion dynamics of the ball based on Gaussian Process Regression equipped with basis functions obtained from physics first principles. The accuracy of this semiparametric model is shown not only in estimation but also in prediction at n-steps ahead and its compared with standard algorithms for model learning. The learned model is then used in a trajectory optimization algorithm to compute ball trajectories. We propose the system presented in the paper as a benchmark problem for reinforcement and robot learning, for its interesting and challenging dynamics and its relative ease of reproducibility.

Jeroen van Baar, Alan Sullivan, Radu Cordorel et al.

Learning robot tasks or controllers using deep reinforcement learning has been proven effective in simulations. Learning in simulation has several advantages. For example, one can fully control the simulated environment, including halting motions while performing computations. Another advantage when robots are involved, is that the amount of time a robot is occupied learning a task---rather than being productive---can be reduced by transferring the learned task to the real robot. Transfer learning requires some amount of fine-tuning on the real robot. For tasks which involve complex (non-linear) dynamics, the fine-tuning itself may take a substantial amount of time. In order to reduce the amount of fine-tuning we propose to learn robustified controllers in simulation. Robustified controllers are learned by exploiting the ability to change simulation parameters (both appearance and dynamics) for successive training episodes. An additional benefit for this approach is that it alleviates the precise determination of physics parameters for the simulator, which is a non-trivial task. We demonstrate our proposed approach on a real setup in which a robot aims to solve a maze game, which involves complex dynamics due to static friction and potentially large accelerations. We show that the amount of fine-tuning in transfer learning for a robustified controller is substantially reduced compared to a non-robustified controller.

Diego Romeres, Mattia Zorzi, Raffaello Camoriano et al.

This paper presents a semi-parametric algorithm for online learning of a robot inverse dynamics model. It combines the strength of the parametric and non-parametric modeling. The former exploits the rigid body dynamics equa- tion, while the latter exploits a suitable kernel function. We provide an extensive comparison with other methods from the literature using real data from the iCub humanoid robot. In doing so we also compare two different techniques, namely cross validation and marginal likelihood optimization, for estimating the hyperparameters of the kernel function.

Diego Romeres, Giulia Prando, Gianluigi Pillonetto et al.

We consider an on-line system identification setting, in which new data become available at given time steps. In order to meet real-time estimation requirements, we propose a tailored Bayesian system identification procedure, in which the hyper-parameters are still updated through Marginal Likelihood maximization, but after only one iteration of a suitable iterative optimization algorithm. Both gradient methods and the EM algorithm are considered for the Marginal Likelihood optimization. We compare this "1-step" procedure with the standard one, in which the optimization method is run until convergence to a local minimum. The experiments we perform confirm the effectiveness of the approach we propose.

Diego Romeres, Gianluigi Pillonetto, Alessandro Chiuso

A new Bayesian approach to linear system identification has been proposed in a series of recent papers. The main idea is to frame linear system identification as predictor estimation in an infinite dimensional space, with the aid of regularization/Bayesian techniques. This approach guarantees the identification of stable predictors based on the prediction error minimization. Unluckily, the stability of the predictors does not guarantee the stability of the impulse response of the system. In this paper we propose and compare various techniques to address this issue. Simulations results comparing these techniques will be provided.