Hamilton-Jacobi Reachability in Reinforcement Learning: A Survey

Recent literature has proposed approaches that learn control policies with high performance while maintaining safety guarantees. Synthesizing Hamilton-Jacobi (HJ) reachable sets has become an effective tool for verifying safety and supervising the training of reinforcement learning-based control policies for complex, high-dimensional systems. Previously, HJ reachability was limited to verifying low-dimensional dynamical systems -- this is because the computational complexity of the dynamic programming approach it relied on grows exponentially with the number of system states. To address this limitation, in recent years, there have been methods that compute the reachability value function simultaneously with learning control policies to scale HJ reachability analysis while still maintaining a reliable estimate of the true reachable set. These HJ reachability approximations are used to improve the safety, and even reward performance, of learned control policies and can solve challenging tasks such as those with dynamic obstacles and/or with lidar-based or vision-based observations. In this survey paper, we review the recent developments in the field of HJ reachability estimation in reinforcement learning that would provide a foundational basis for further research into reliability in high-dimensional systems.

Mollification Effects of Policy Gradient Methods

Policy gradient methods have enabled deep reinforcement learning (RL) to approach challenging continuous control problems, even when the underlying systems involve highly nonlinear dynamics that generate complex non-smooth optimization landscapes. We develop a rigorous framework for understanding how policy gradient methods mollify non-smooth optimization landscapes to enable effective policy search, as well as the downside of it: while making the objective function smoother and easier to optimize, the stochastic objective deviates further from the original problem. We demonstrate the equivalence between policy gradient methods and solving backward heat equations. Following the ill-posedness of backward heat equations from PDE theory, we present a fundamental challenge to the use of policy gradient under stochasticity. Moreover, we make the connection between this limitation and the uncertainty principle in harmonic analysis to understand the effects of exploration with stochastic policies in RL. We also provide experimental results to illustrate both the positive and negative aspects of mollification effects in practice.

Sensor-Based Distributionally Robust Control for Safe Robot Navigation in Dynamic Environments

We introduce a novel method for safe mobile robot navigation in dynamic, unknown environments, utilizing onboard sensing to impose safety constraints without the need for accurate map reconstruction. Traditional methods typically rely on detailed map information to synthesize safe stabilizing controls for mobile robots, which can be computationally demanding and less effective, particularly in dynamic operational conditions. By leveraging recent advances in distributionally robust optimization, we develop a distributionally robust control barrier function (DR-CBF) constraint that directly processes range sensor data to impose safety constraints. Coupling this with a control Lyapunov function (CLF) for path tracking, we demonstrate that our CLF-DR-CBF control synthesis method achieves safe, efficient, and robust navigation in uncertain dynamic environments. We demonstrate the effectiveness of our approach in simulated and real autonomous robot navigation experiments, marking a substantial advancement in real-time safety guarantees for mobile robots.

JIGGLE: An Active Sensing Framework for Boundary Parameters Estimation in Deformable Surgical Environments

Surgical automation can improve the accessibility and consistency of life saving procedures. Most surgeries require separating layers of tissue to access the surgical site, and suturing to reattach incisions. These tasks involve deformable manipulation to safely identify and alter tissue attachment (boundary) topology. Due to poor visual acuity and frequent occlusions, surgeons tend to carefully manipulate the tissue in ways that enable inference of the tissue's attachment points without causing unsafe tearing. In a similar fashion, we propose JIGGLE, a framework for estimation and interactive sensing of unknown boundary parameters in deformable surgical environments. This framework has two key components: (1) a probabilistic estimation to identify the current attachment points, achieved by integrating a differentiable soft-body simulator with an extended Kalman filter (EKF), and (2) an optimization-based active control pipeline that generates actions to maximize information gain of the tissue attachments, while simultaneously minimizing safety costs. The robustness of our estimation approach is demonstrated through experiments with real animal tissue, where we infer sutured attachment points using stereo endoscope observations. We also demonstrate the capabilities of our method in handling complex topological changes such as cutting and suturing.

Parameterized Fast and Safe Tracking (FaSTrack) using Deepreach

Fast and Safe Tracking (FaSTrack) is a modular framework that provides safety guarantees while planning and executing trajectories in real time via value functions of Hamilton-Jacobi (HJ) reachability. These value functions are computed through dynamic programming, which is notorious for being computationally inefficient. Moreover, the resulting trajectory does not adapt online to the environment, such as sudden disturbances or obstacles. DeepReach is a scalable deep learning method to HJ reachability that allows parameterization of states, which opens up possibilities for online adaptation to various controls and disturbances. In this paper, we propose Parametric FaSTrack, which uses DeepReach to approximate a value function that parameterizes the control bounds of the planning model. The new framework can smoothly trade off between the navigation speed and the tracking error (therefore maneuverability) while guaranteeing obstacle avoidance in a priori unknown environments. We demonstrate our method through two examples and a benchmark comparison with existing methods, showing the safety, efficiency, and faster solution times of the framework.

Safe Returning FaSTrack with Robust Control Lyapunov-Value Functions

Real-time navigation in a priori unknown environment remains a challenging task, especially when an unexpected (unmodeled) disturbance occurs. In this paper, we propose the framework Safe Returning Fast and Safe Tracking (SR-F) that merges concepts from 1) Robust Control Lyapunov-Value Functions (R-CLVF), and 2) the Fast and Safe Tracking (FaSTrack) framework. The SR-F computes an R-CLVF offline between a model of the true system and a simplified planning model. Online, a planning algorithm is used to generate a trajectory in the simplified planning space, and the R-CLVF is used to provide a tracking controller that exponentially stabilizes to the planning model. When an unexpected disturbance occurs, the proposed SR-F algorithm provides a means for the true system to recover to the planning model. We take advantage of this mechanism to induce an artificial disturbance by ``jumping'' the planning model in open environments, forcing faster navigation. Therefore, this algorithm can both reject unexpected true disturbances and accelerate navigation speed. We validate our framework using a 10D quadrotor system and show that SR-F is empirically 20\% faster than the original FaSTrack while maintaining safety.

SURESTEP: An Uncertainty-Aware Trajectory Optimization Framework to Enhance Visual Tool Tracking for Robust Surgical Automation

Inaccurate tool localization is one of the main reasons for failures in automating surgical tasks. Imprecise robot kinematics and noisy observations caused by the poor visual acuity of an endoscopic camera make tool tracking challenging. Previous works in surgical automation adopt environment-specific setups or hard-coded strategies instead of explicitly considering motion and observation uncertainty of tool tracking in their policies. In this work, we present SURESTEP, an uncertainty-aware trajectory optimization framework for robust surgical automation. We model the uncertainty of tool tracking with the components motivated by the sources of noise in typical surgical scenes. Using a Gaussian assumption to propagate our uncertainty models through a given tool trajectory, SURESTEP provides a general framework that minimizes the upper bound on the entropy of the final estimated tool distribution. We compare SURESTEP with a baseline method on a real-world suture needle regrasping task under challenging environmental conditions, such as poor lighting and a moving endoscopic camera. The results over 60 regrasps on the da Vinci Research Kit (dVRK) demonstrate that our optimized trajectories significantly outperform the un-optimized baseline.

Fractal Landscapes in Policy Optimization

Policy gradient lies at the core of deep reinforcement learning (RL) in continuous domains. Despite much success, it is often observed in practice that RL training with policy gradient can fail for many reasons, even on standard control problems with known solutions. We propose a framework for understanding one inherent limitation of the policy gradient approach: the optimization landscape in the policy space can be extremely non-smooth or fractal for certain classes of MDPs, such that there does not exist gradient to be estimated in the first place. We draw on techniques from chaos theory and non-smooth analysis, and analyze the maximal Lyapunov exponents and H\"older exponents of the policy optimization objectives. Moreover, we develop a practical method that can estimate the local smoothness of objective function from samples to identify when the training process has encountered fractal landscapes. We show experiments to illustrate how some failure cases of policy optimization can be explained by such fractal landscapes.

Iterative Reachability Estimation for Safe Reinforcement Learning

Ensuring safety is important for the practical deployment of reinforcement learning (RL). Various challenges must be addressed, such as handling stochasticity in the environments, providing rigorous guarantees of persistent state-wise safety satisfaction, and avoiding overly conservative behaviors that sacrifice performance. We propose a new framework, Reachability Estimation for Safe Policy Optimization (RESPO), for safety-constrained RL in general stochastic settings. In the feasible set where there exist violation-free policies, we optimize for rewards while maintaining persistent safety. Outside this feasible set, our optimization produces the safest behavior by guaranteeing entrance into the feasible set whenever possible with the least cumulative discounted violations. We introduce a class of algorithms using our novel reachability estimation function to optimize in our proposed framework and in similar frameworks such as those concurrently handling multiple hard and soft constraints. We theoretically establish that our algorithms almost surely converge to locally optimal policies of our safe optimization framework. We evaluate the proposed methods on a diverse suite of safe RL environments from Safety Gym, PyBullet, and MuJoCo, and show the benefits in improving both reward performance and safety compared with state-of-the-art baselines.

Object-centric Representations for Interactive Online Learning with Non-Parametric Methods

Large offline learning-based models have enabled robots to successfully interact with objects for a wide variety of tasks. However, these models rely on fairly consistent structured environments. For more unstructured environments, an online learning component is necessary to gather and estimate information about objects in the environment in order to successfully interact with them. Unfortunately, online learning methods like Bayesian non-parametric models struggle with changes in the environment, which is often the desired outcome of interaction-based tasks. We propose using an object-centric representation for interactive online learning. This representation is generated by transforming the robot's actions into the object's coordinate frame. We demonstrate how switching to this task-relevant space improves our ability to reason with the training data collected online, enabling scalable online learning of robot-object interactions. We showcase our method by successfully navigating a manipulator arm through an environment with multiple unknown objects without violating interaction-based constraints.