Maximilian Xiling Li

Maximilian Xiling Li, Onur Celik, Philipp Becker et al.

Learning skills by imitation is a promising concept for the intuitive teaching of robots. A common way to learn such skills is to learn a parametric model by maximizing the likelihood given the demonstrations. Yet, human demonstrations are often multi-modal, i.e., the same task is solved in multiple ways which is a major challenge for most imitation learning methods that are based on such a maximum likelihood (ML) objective. The ML objective forces the model to cover all data, it prevents specialization in the context space and can cause mode-averaging in the behavior space, leading to suboptimal or potentially catastrophic behavior. Here, we alleviate those issues by introducing a curriculum using a weight for each data point, allowing the model to specialize on data it can represent while incentivizing it to cover as much data as possible by an entropy bonus. We extend our algorithm to a Mixture of (linear) Experts (MoE) such that the single components can specialize on local context regions, while the MoE covers all data points. We evaluate our approach in complex simulated and real robot control tasks and show it learns from versatile human demonstrations and significantly outperforms current SOTA methods. A reference implementation can be found at https://github.com/intuitive-robots/ml-cur

Mohammad Taufeeque, Philip Quirke, Maximilian Li et al.

Planning is essential for solving complex tasks, yet the internal mechanisms underlying planning in neural networks remain poorly understood. Building on prior work, we analyze a recurrent neural network (RNN) trained on Sokoban, a challenging puzzle requiring sequential, irreversible decisions. We find that the RNN has a causal plan representation which predicts its future actions about 50 steps in advance. The quality and length of the represented plan increases over the first few steps. We uncover a surprising behavior: the RNN "paces" in cycles to give itself extra computation at the start of a level, and show that this behavior is incentivized by training. Leveraging these insights, we extend the trained RNN to significantly larger, out-of-distribution Sokoban puzzles, demonstrating robust representations beyond the training regime. We open-source our model and code, and believe the neural network's interesting behavior makes it an excellent model organism to deepen our understanding of learned planning.

Moritz Reuss, Maximilian Li, Xiaogang Jia et al.

We propose a new policy representation based on score-based diffusion models (SDMs). We apply our new policy representation in the domain of Goal-Conditioned Imitation Learning (GCIL) to learn general-purpose goal-specified policies from large uncurated datasets without rewards. Our new goal-conditioned policy architecture "$\textbf{BE}$havior generation with $\textbf{S}$c$\textbf{O}$re-based Diffusion Policies" (BESO) leverages a generative, score-based diffusion model as its policy. BESO decouples the learning of the score model from the inference sampling process, and, hence allows for fast sampling strategies to generate goal-specified behavior in just 3 denoising steps, compared to 30+ steps of other diffusion based policies. Furthermore, BESO is highly expressive and can effectively capture multi-modality present in the solution space of the play data. Unlike previous methods such as Latent Plans or C-Bet, BESO does not rely on complex hierarchical policies or additional clustering for effective goal-conditioned behavior learning. Finally, we show how BESO can even be used to learn a goal-independent policy from play-data using classifier-free guidance. To the best of our knowledge this is the first work that a) represents a behavior policy based on such a decoupled SDM b) learns an SDM based policy in the domain of GCIL and c) provides a way to simultaneously learn a goal-dependent and a goal-independent policy from play-data. We evaluate BESO through detailed simulation and show that it consistently outperforms several state-of-the-art goal-conditioned imitation learning methods on challenging benchmarks. We additionally provide extensive ablation studies and experiments to demonstrate the effectiveness of our method for goal-conditioned behavior generation. Demonstrations and Code are available at https://intuitive-robots.github.io/beso-website/

Paul Werner Lödige, Maximilian Xiling Li, Rudolf Lioutikov

Movement Primitives (MPs) are a well-established method for representing and generating modular robot trajectories. This work presents FA-ProDMP, a new approach which introduces force awareness to Probabilistic Dynamic Movement Primitives (ProDMP). FA-ProDMP adapts the trajectory during runtime to account for measured and desired forces. It offers smooth trajectories and captures position and force correlations over multiple trajectories, e.g. a set of human demonstrations. FA-ProDMP supports multiple axes of force and is thus agnostic to cartesian or joint space control. This makes FA-ProDMP a valuable tool for learning contact rich manipulation tasks such as polishing, cutting or industrial assembly from demonstration. In order to reliably evaluate FA-ProDMP, this work additionally introduces a modular, 3D printed task suite called POEMPEL, inspired by the popular Lego Technic pins. POEMPEL mimics industrial peg-in-hole assembly tasks with force requirements. It offers multiple parameters of adjustment, such as position, orientation and plug stiffness level, thus varying the direction and amount of required forces. Our experiments show that FA-ProDMP outperforms other MP formulations on the POEMPEL setup and a electrical power plug insertion task, due to its replanning capabilities based on the measured forces. These findings highlight how FA-ProDMP enhances the performance of robotic systems in contact-rich manipulation tasks.

Maximilian Li, Xander Davies, Max Nadeau

Language models often exhibit behaviors that improve performance on a pre-training objective but harm performance on downstream tasks. We propose a novel approach to removing undesirable behaviors by ablating a small number of causal pathways between model components, with the intention of disabling the computational circuit responsible for the bad behavior. Given a small dataset of inputs where the model behaves poorly, we learn to ablate a small number of important causal pathways. In the setting of reducing GPT-2 toxic language generation, we find ablating just 12 of the 11.6K causal edges mitigates toxic generation with minimal degradation of performance on other inputs.

Denis Blessing, Onur Celik, Xiaogang Jia et al.

Imitation learning uses data for training policies to solve complex tasks. However, when the training data is collected from human demonstrators, it often leads to multimodal distributions because of the variability in human actions. Most imitation learning methods rely on a maximum likelihood (ML) objective to learn a parameterized policy, but this can result in suboptimal or unsafe behavior due to the mode-averaging property of the ML objective. In this work, we propose Information Maximizing Curriculum, a curriculum-based approach that assigns a weight to each data point and encourages the model to specialize in the data it can represent, effectively mitigating the mode-averaging problem by allowing the model to ignore data from modes it cannot represent. To cover all modes and thus, enable diverse behavior, we extend our approach to a mixture of experts (MoE) policy, where each mixture component selects its own subset of the training data for learning. A novel, maximum entropy-based objective is proposed to achieve full coverage of the dataset, thereby enabling the policy to encompass all modes within the data distribution. We demonstrate the effectiveness of our approach on complex simulated control tasks using diverse human demonstrations, achieving superior performance compared to state-of-the-art methods.

Maximilian Xiling Li, Korbinian Franz Rudolf, Nils Blank et al.

Prototypical part networks offer interpretable alternatives to black-box deep learning models. However, many of these networks rely on Euclidean prototypes, which may limit their flexibility. This work provides a comprehensive overview of various prototype formulations. Experiments conducted on the CUB-200-2011, Stanford Cars, and Oxford Flowers datasets demonstrate the effectiveness and versatility of these different formulations.

Maximilian Xiling Li, Korbinian Rudolf, Nils Blank et al.

Robotic agents need to understand how to interact with objects in their environment, both autonomously and during human-robot interactions. Affordance detection on 3D point clouds, which identifies object regions that allow specific interactions, has traditionally relied on deep learning models like PointNet++, DGCNN, or PointTransformerV3. However, these models operate as black boxes, offering no insight into their decision-making processes. Prototypical Learning methods, such as ProtoPNet, provide an interpretable alternative to black-box models by employing a "this looks like that" case-based reasoning approach. However, they have been primarily applied to image-based tasks. In this work, we apply prototypical learning to models for affordance detection on 3D point clouds. Experiments on the 3D-AffordanceNet benchmark dataset show that prototypical models achieve competitive performance with state-of-the-art black-box models and offer inherent interpretability. This makes prototypical models a promising candidate for human-robot interaction scenarios that require increased trust and safety.