Probabilistically Safe Robot Planning with Confidence-Based Human Predictions

Abstract

Abstract-In order to safely operate around humans, robots can employ predictive models of human motion. Unfortunately, these models cannot capture the full complexity of human behavior and necessarily introduce simplifying assumptions. As a result, predictions may degrade whenever the observed human behavior departs from the assumed structure, which can have negative implications for safety. In this paper, we observe that how "rational" human actions appear under a particular model can be viewed as an indicator of that model's ability to describe the human's current motion. By reasoning about this model confidence in a real-time Bayesian framework, we show that the robot can very quickly modulate its predictions to become more uncertain when the model performs poorly. Building on recent work in provably-safe trajectory planning, we leverage these confidence-aware human motion predictions to generate assured autonomous robot motion. Our new analysis combines worst-case tracking error guarantees for the physical robot with probabilistic time-varying human predictions, yielding a quantitative, probabilistic safety certificate. We demonstrate our approach with a quadcopter navigating around a human.

Quantifying Hypothesis Space Misspecification in Learning From Human–Robot Demonstrations and Physical Corrections

“…To discretize situational confidence, we found it sufficient to cover β ∈ {0.01, 0.03, 0.1, 0.3, 1.0, 3.0, 10.0, 30.0, 100.0}, the log-scale space, similarly to [30], [31]. For different tasks, a similar discretization should suffice because what matters is β's relative magnitude for identifying misspecification, not its absolute one.…”

“…0), it returns a reasonable estimate of the true u * H . (24) and (30) rely heavily on hyperparameters λ and ν. Here, we discuss how to set them.…”

Efficient Probabilistic Collision Detection for Non-Gaussian Noise Distributions

“…The upper bound on collision probability is computed on the Gaussian positional error with an erf(·) function for a point obstacle. Fisac et al [24] compute the collision probability between the dynamic human motion and a robot, and use that value for robot motion planning in the 3D workspace. This algorithm models the human motion based on human dynamics, discretizes the 3D workspace into smaller grids, and integrates the cell probabilities over the volume occupied by the robot.…”

A Scalable Framework For Real-Time Multi-Robot, Multi-Human Collision Avoidance

“…Recent work, however, has made progress in provably-safe real-time motion planning [3][4][5] 1 Note that our laboratory setting uses a motion capture system for sensing and state estimation-robustness with respect to sensor uncertainty is an important component that is beyond the scope of this paper. probabilistic prediction of a human agent's motion [6,7], and robust sequential trajectory planning for multi-robot systems [8,9]. It remains a challenge to synthesize these into a real-time planning system, primarily due to the difficulty of joint planning and prediction for multiple robots and humans.…”

“…To ensure real-time feasibility, robots predict human motion using a simple model neglecting future interaction effects. Because this model will be a simplification of true human motion, we use confidence-aware predictions [6] that become more conservative whenever humans deviate from the assumed model. Finally, groups of robots plan sequentially according to a pre-specified priority ordering [13], which serves to reduce the complexity of the joint planning problem while maintaining safety with respect to each other.…”