Neural networks improve real-time robot safety in changing environments
Learning Reachability-Based Safety Representations for Real-Time Robot Navigation in Unknown and Dynamic Environments
RoboticsMachine Learning
Summary
Robots moving around unfamiliar and moving obstacles need to be very careful to avoid crashes. The paper’s author uses math to identify states where robots can safely continue moving and trains neural networks to quickly predict this in real time. Their approach helps robots better avoid collisions by considering factors like speed and obstacles' movements, not just distance. Tests on robots on the ground and in the air showed this method helps avoid crashes better than simpler methods.
What this means in practice
- •For robotics engineers: Improve safety in robot navigation systems by integrating learned reachability-based safety checks for real-time decision making in complex, dynamic environments.
- •For drone fleet operators: Enhance collision avoidance algorithms for aerial drones handling moving obstacles by using neural network informed control barrier functions.
Authors
Bojan Derajić
Abstract
Safe robot navigation in unknown, dynamic environments requires more than geometric collision checking: a robot still far from an obstacle may be unable to avoid it because of its speed, orientation, or the obstacle's motion. Hamilton--Jacobi (HJ) reachability analysis captures this in a value function whose sign separates states that admit a safe continuation from those where collision is unavoidable, but computing it online for every newly observed environment is too expensive. This dissertation trains neural implicit representations offline on reachability solutions to give optimization-based controllers this information in real time. Since the value function never exceeds the constraint, it equals a directly computed constraint, such as a signed distance, minus a nonnegative dynamics correction. Learning only this correction keeps represented obstacles outside the predicted safe set despite approximation errors. The first two of five methods use the learned function as a terminal constraint in model predictive control, for static local environments and predicted obstacle motion. The next two use it as a barrier candidate in control barrier function safety filters, for static environments and, through pairwise barriers in relative coordinates, for a varying number of moving obstacles. Simulations and hardware experiments on ground and aerial robots show better collision avoidance than geometric and hand-designed baselines, especially for short horizons and moving obstacles. The fifth method replaces the grid-based offline solver, which limits the first four to low-dimensional models. For control-affine dynamics with polytopic inputs, the HJ Hamiltonian is maximized at a vertex, so a parallelizable tree search over vertex inputs can generate training labels. With a loss from the HJ variational inequality, these labels yield barrier candidates for systems with up to 13 states.