Search papers, labs, and topics across Lattice.
This paper introduces a risk-aware formalism for evaluating system trajectories, explicitly modeling the influence of trajectories on the environment and evaluating them under the resulting distribution of environment responses. The formalism supports reasoning under uncertainty, handles complex relationships among requirements and objectives, and induces a preorder on the set of system trajectories, ensuring consistency. The approach is demonstrated in an autonomous driving example, enhancing explainability in trajectory selection.
Autonomous systems can now make more explainable and consistent decisions under uncertainty by explicitly modeling how their actions influence the environment.
We present a risk-aware formalism for evaluating system trajectories in the presence of uncertain interactions between the system and its environment. The proposed formalism supports reasoning under uncertainty and systematically handles complex relationships among requirements and objectives, including hierarchical priorities and non-comparability. Rather than treating the environment as exogenous noise, we explicitly model how each system trajectory influences the environment and evaluate trajectories under the resulting distribution of environment responses. We prove that the formalism induces a preorder on the set of system trajectories, ensuring consistency and preventing cyclic preferences. Finally, we illustrate the approach with an autonomous driving example that demonstrates how the formalism enhances explainability by clarifying the rationale behind trajectory selection.
