Search papers, labs, and topics across Lattice.
This paper introduces a risk-aware belief control barrier function (BCBF) framework to ensure robot safety in dynamic environments with an unknown number of moving objects. By leveraging a random finite set (RFS) belief estimated through a sequential Monte Carlo probability hypothesis density (SMC-PHD) filter, the authors construct a nonsmooth BCBF that guarantees forward invariance of the safe set and establishes conditions for safety preservation during discrete updates. Experimental results in both simulations and real-world underwater scenarios validate the framework's effectiveness and efficiency in managing multi-object state uncertainty.
Robots can now navigate safely in unpredictable environments by effectively managing multi-object uncertainties with a novel risk-aware control framework.
Ensuring robot safety in unknown, dynamic environments is a fundamental requirement. It involves inferring the states of an unknown and time-varying number of moving objects from noisy, incomplete measurements. We address safe control under the induced multi-object state uncertainty with a risk-aware belief control barrier function (BCBF) framework. The uncertainty is captured by a random finite set (RFS) belief, estimated by a sequential Monte Carlo probability hypothesis density (SMC-PHD) filter that represents it with a set of particles. Building directly on these particles, we construct a nonsmooth BCBF, establish forward invariance of the safe set under continuous prediction, and derive an explicit condition under which discrete updates preserve safety. Simulation and real-world underwater experiments demonstrate the effectiveness and efficiency of the proposed approach.
