With the increasing availability of open-source robotic data, imitation learning has emerged as a viable approach for both robot manipulation and locomotion. Currently, large generalized policies are trained to predict controls or trajectories using diffusion models, which have the desirable property of learning multimodal action distributions. However, generalizability comes with a cost - namely, larger model size and slower inference. Further, there is a known trade-off between performance and action horizon for Diffusion Policy (i.e., diffusing trajectories): fewer diffusion queries accumulate greater trajectory tracking errors. Thus, it is common practice to run these models at high inference frequency, subject to robot computational constraints. To address these limitations, we propose Latent Weight Diffusion (LWD), a method that uses diffusion to learn a distribution over policies for robotic tasks, rather than over trajectories. Our approach encodes demonstration trajectories into a latent space and then decodes them into policies using a hypernetwork. We employ a diffusion denoising model within this latent space to learn its distribution. We demonstrate that LWD can reconstruct the behaviors of the original policies that generated the trajectory dataset. LWD offers the benefits of considerably smaller policy networks during inference and requires fewer diffusion model queries. When tested on the Metaworld MT10 benchmark, LWD achieves a higher success rate compared to a vanilla multi-task policy, while using models up to ~18x smaller during inference. Additionally, since LWD generates closed-loop policies, we show that it outperforms Diffusion Policy in long action horizon settings, with reduced diffusion queries during rollout.

Generalizing vision-based reinforcement learning (RL) agents to novel environments remains a difficult and open challenge. Current trends are to collect large-scale datasets or use data augmentation techniques to prevent overfitting and improve downstream generalization. However, the computational and data collection costs increase exponentially with the number of task variations and can destabilize the already difficult task of training RL agents. In this work, we take inspiration from recent advances in computational neuroscience and propose a model, Associative Latent DisentAnglement (ALDA), that builds on standard off-policy RL towards zero-shot generalization. Specifically, we revisit the role of latent disentanglement in RL and show how combining it with a model of associative memory achieves zero-shot generalization on difficult task variations without relying on data augmentation. Finally, we formally show that data augmentation techniques are a form of weak disentanglement and discuss the implications of this insight.

Robotic assembly for high-mixture settings requires adaptivity to diverse parts and poses, which is an open challenge. Meanwhile, in other areas of robotics, large models and sim-to-real have led to tremendous progress. Inspired by such work, we present AutoMate, a learning framework and system that consists of 4 parts: 1) a dataset of 100 assemblies compatible with simulation and the real world, along with parallelized simulation environments for policy learning, 2) a novel simulation-based approach for learning specialist (i.e., part-specific) policies and generalist (i.e., unified) assembly policies, 3) demonstrations of specialist policies that individually solve 80 assemblies with 80% or higher success rates in simulation, as well as a generalist policy that jointly solves 20 assemblies with an 80%+ success rate, and 4) zero-shot sim-to-real transfer that achieves similar (or better) performance than simulation, including on perception-initialized assembly. The key methodological takeaway is that a union of diverse algorithms from manufacturing engineering, character animation, and time-series analysis provides a generic and robust solution for a diverse range of robotic assembly problems.To our knowledge, AutoMate provides the first simulation-based framework for learning specialist and generalist policies over a wide range of assemblies, as well as the first system demonstrating zero-shot sim-to-real transfer over such a range.

Multi-robot target tracking finds extensive applications in different scenarios, such as environmental surveillance and wildfire management, which require the robustness of the practical deployment of multi-robot systems in uncertain and dangerous environments. Traditional approaches often focus on the performance of tracking accuracy with no modeling and assumption of the environments, neglecting potential environmental hazards which result in system failures in real-world deployments. To address this challenge, we investigate multi-robot target tracking in the adversarial environment considering sensing and communication attacks with uncertainty. We design specific strategies to avoid different danger zones and proposed a multi-agent tracking framework under the perilous environment. We approximate the probabilistic constraints and formulate practical optimization strategies to address computational challenges efficiently. We evaluate the performance of our proposed methods in simulations to demonstrate the ability of robots to adjust their risk-aware behaviors under different levels of environmental uncertainty and risk confidence. The proposed method is further validated via real-world robot experiments where a team of drones successfully track dynamic ground robots while being risk-aware of the sensing and/or communication danger zones.

We consider a new variant of the multi-robot task allocation problem - Inverse Risk-sensitive Multi-Robot Task Allocation (IR-MRTA). "Forward" MRTA - the process of deciding which robot should perform a task given the reward (cost)-related parameters, is widely studied in the multi-robot literature. In this setting, the reward (cost)-related parameters are assumed to be already known: parameters are first fixed offline by domain experts, followed by coordinating robots online. What if we need these parameters to be adjusted by non-expert human supervisors who oversee the robots during tasks to adapt to new situations? We are interested in the case where the human supervisor's perception of the allocation risk may change and suggest different allocations for robots compared to that from the MRTA algorithm. In such cases, the robots need to change the parameters of the allocation problem based on evolving human preferences. We study such problems through the lens of inverse task allocation, i.e., the process of finding parameters given solutions to the problem. Specifically, we propose a new formulation IR-MRTA, in which we aim to find a new set of parameters of the human behavioral risk model that minimally deviates from the current MRTA parameters and can make a greedy task allocation algorithm allocate robot resources in line with those suggested by humans. We show that even in the simple case such a problem is a non-convex optimization problem. We propose a Branch $\&$ Bound algorithm (BB-IR-MRTA) to solve such problems. In numerical simulations of a case study on multi-robot target capture, we demonstrate how to use BB-IR-MRTA and we show that the proposed algorithm achieves significant advantages in running time and peak memory usage compared to a brute-force baseline.

Placement of electromagnetic signal emitting devices, such as light sources, has important usage in for signal coverage tasks. Automatic placement of these devices is challenging because of the complex interaction of the signal and environment due to reflection, refraction and scattering. In this work, we iteratively improve the placement of these devices by interleaving device placement and sensing actions, correcting errors in the model of the signal propagation. To this end, we propose a novel factor-graph based belief model which combines the measurements taken by the robot and an analytical light propagation model. This model allows accurately modelling the uncertainty of the light propagation with respect to the obstacles, which greatly improves the informative path planning routine. Additionally, we propose a method for determining when to re-plan the emitter placements to balance a trade-off between information about a specific configuration and frequent updating of the configuration. This method incorporates the uncertainty from belief model to adaptively determine when re-configuration is needed. We find that our system has a 9.8% median error reduction compared to a baseline system in simulations in the most difficult environment. We also run on-robot tests and determine that our system performs favorably compared to the baseline.

We consider a new type of inverse combinatorial optimization, Inverse Submodular Maximization (ISM), for human-in-the-loop multi-robot coordination. Forward combinatorial optimization, defined as the process of solving a combinatorial problem given the reward (cost)-related parameters, is widely used in multi-robot coordination. In the standard pipeline, the reward (cost)-related parameters are designed offline by domain experts first and then these parameters are utilized for coordinating robots online. What if we need to change these parameters by non-expert human supervisors who watch over the robots during tasks to adapt to some new requirements? We are interested in the case where human supervisors can suggest what actions to take, and the robots need to change the internal parameters based on such suggestions. We study such problems from the perspective of inverse combinatorial optimization, i.e., the process of finding parameters given solutions to the problem. Specifically, we propose a new formulation for ISM, in which we aim to find a new set of parameters that minimally deviate from the current parameters and can make the greedy algorithm output actions the same as those suggested by humans. We show that such problems can be formulated as a Mixed Integer Quadratic Program (MIQP). However, MIQP involves exponentially many binary variables, making it intractable for the existing solver when the problem size is large. We propose a new algorithm under the Branch $\&$ Bound paradigm to solve such problems. In numerical simulations, we demonstrate how to use ISM in multi-robot multi-objective coverage control, and we show that the proposed algorithm achieves significant advantages in running time and peak memory usage compared to directly using an existing solver.

LLMs have shown impressive progress in robotics (e.g., manipulation and navigation) with natural language task descriptions. The success of LLMs in these tasks leads us to wonder: What is the ability of LLMs to solve vehicle routing problems (VRPs) with natural language task descriptions? In this work, we study this question in three steps. First, we construct a dataset with 21 types of single- or multi-vehicle routing problems. Second, we evaluate the performance of LLMs across four basic prompt paradigms of text-to-code generation, each involving different types of text input. We find that the basic prompt paradigm, which generates code directly from natural language task descriptions, performs the best for GPT-4, achieving 56% feasibility, 40% optimality, and 53% efficiency. Third, based on the observation that LLMs may not be able to provide correct solutions at the initial attempt, we propose a framework that enables LLMs to refine solutions through self-reflection, including self-debugging and self-verification. With GPT-4, our proposed framework achieves a 16% increase in feasibility, a 7% increase in optimality, and a 15% increase in efficiency. Moreover, we examine the sensitivity of GPT-4 to task descriptions, specifically focusing on how its performance changes when certain details are omitted from the task descriptions, yet the core meaning is preserved. Our findings reveal that such omissions lead to a notable decrease in performance: 4% in feasibility, 4% in optimality, and 5% in efficiency. Website: https://sites.google.com/view/words-to-routes/

Large, high-capacity models trained on diverse datasets have shown remarkable successes on efficiently tackling downstream applications. In domains from NLP to Computer Vision, this has led to a consolidation of pretrained models, with general pretrained backbones serving as a starting point for many applications. Can such a consolidation happen in robotics? Conventionally, robotic learning methods train a separate model for every application, every robot, and even every environment. Can we instead train generalist X-robot policy that can be adapted efficiently to new robots, tasks, and environments? In this paper, we provide datasets in standardized data formats and models to make it possible to explore this possibility in the context of robotic manipulation, alongside experimental results that provide an example of effective X-robot policies. We assemble a dataset from 22 different robots collected through a collaboration between 21 institutions, demonstrating 527 skills (160266 tasks). We show that a high-capacity model trained on this data, which we call RT-X, exhibits positive transfer and improves the capabilities of multiple robots by leveraging experience from other platforms. More details can be found on the project website $\href{https://robotics-transformer-x.github.io}{\text{robotics-transformer-x.github.io}}$.

On-policy reinforcement learning (RL) methods seek to optimize a stochastic policy, where a neural network is used to parameterize a distribution π(a|s) over actions conditioned on the current state. In this framework, most on-policy RL methods seek to limit the scale of updates between successive policies during optimization. Some on-policy RL methods operate by guaranteeing that each policy update remains within a "trust region" (Schulman et al., 2015a). These methods are used when training stability during a long period of training is essential. However, finding a policy update near the edge of the trust region often comes at significant computational cost. Another branch of on-policy methods instead perform "proximal" policy updates, that limit the expected scale of policy updates, but can result in individual policy updates being of arbitrary magnitude (Schulman et al., 2017a). These methods are much more computationally efficient, but large-scale training can require the use of multiple training runs or human intervention to recover from training instabilities. In this work we propose Fixup Policy Optimization (FixPO), which combines both a proximal primary phase with a precise fixup phase, that operate by sharing a single penalty coefficient β. By performing a more conservative proximal update before strictly enforcing a trust region, FixPO is able to approximately match the computational efficiency and rewards of proximal methods while providing the same stability guarantees as trust region methods.