Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces pdSTL, a novel framework that integrates probabilistic semantics with differentiable robustness for Signal Temporal Logic (STL) to enhance the performance of autonomous robots in uncertain environments. By employing interval-valued probabilistic semantics and a recurrent, LSTM-style unfolding of STL operators, pdSTL allows for efficient, linear-time monitoring and optimization of trajectories while ensuring formal probabilistic guarantees. Experimental results show that pdSTL significantly outperforms traditional deterministic STL in maintaining safety margins during complex maneuvers under real-world conditions.
pdSTL achieves efficient trajectory optimization with formal probabilistic guarantees, outperforming traditional methods in safety under uncertainty.
Autonomous robots operating in uncertain environments must satisfy complex temporal and safety specifications despite stochastic dynamics and sensing noise. While Signal Temporal Logic (STL) offers robustness measures for gradient-based optimization, existing extensions either lack differentiability or ignore belief-space uncertainty. We introduce pdSTL (probabilistic differentiable Signal Temporal Logic), a framework that unifies probabilistic semantics with differentiable robustness over belief trajectories. pdSTL employs interval-valued probabilistic semantics to compute conservative satisfaction bounds, propagated compositionally through the STL syntax tree. We formulate the temporal robustness evaluation as a recurrent, LSTM-style unfolding of STL operators, enabling linear-time, differentiable monitoring suitable for end-to-end trajectory optimization. We validate pdSTL on simulated obstacle avoidance, lane-change maneuvers, and real-world Crazyflie quadcopter flight experiments under aerodynamic disturbances. Results demonstrate that pdSTL achieves efficient optimization with formal probabilistic guarantees, significantly outperforming deterministic differentiable STL in maintaining safety margins under real-world uncertainty.