Towards Probabilistic Clearance, Explanation and Optimization

TL;DR

The work addresses safe BVLOS UAS navigation under regulatory constraints and uncertain, crowd-sourced map data. It extends the Probabilistic Mission Design (ProMis) framework by turning Probabilistic Mission Landscapes (PML) into navigation graphs and introducing a clearance–explanation–optimization (CEO) cycle. Key contributions include trajectory planning over PMLs, interpretable explanations of parameter effects on clearance, and an optimization loop that tunes mission settings to maximize constraint compliance, with an Open Source implementation. Overall, the approach yields explainable, risk-aware navigation suitable for urban air mobility and UTM-like systems, while enabling iterative, user-guided mission design under uncertainty.

Abstract

Employing Unmanned Aircraft Systems (UAS) beyond visual line of sight (BVLOS) is an endearing and challenging task. While UAS have the potential to significantly enhance today's logistics and emergency response capabilities, unmanned flying objects above the heads of unprotected pedestrians induce similarly significant safety risks. In this work, we make strides towards improved safety and legal compliance in applying UAS in two ways. First, we demonstrate navigation within the Probabilistic Mission Design (ProMis) framework. To this end, our approach translates Probabilistic Mission Landscapes (PML) into a navigation graph and derives a cost from the probability of complying with all underlying constraints. Second, we introduce the clearance, explanation, and optimization (CEO) cycle on top of ProMis by leveraging the declaratively encoded domain knowledge, legal requirements, and safety assertions to guide the mission design process. Based on inaccurate, crowd-sourced map data and a synthetic scenario, we illustrate the application and utility of our methods in UAS navigation.

Figures (5)

• Figure 1: The probabilistic CEO cycle: If an intended Probabilistic Mission (ProMis) does not pass the clearance (C) step, explanation (E) identifies critical parameters that are passed to an optimizer (O) choosing the ideal setting, e.g., consisting of departure time, operator attributes, or UAS type. This cycle is transparent to the user, only acting on inputs likely to violate legal and safety constraints or operator preferences.
• Figure 2: Hybrid relational parameters and Probabilistic Mission Landscape: Mean (top-left) and variance (top-right) of distance to nearest road and probability of occupation (bottom-left) parameterize the hybrid relational, logical model for defining a Probabilistic Mission Landscape (bottom-right).
• Figure 3: The Probabilistic UAS architecture: Perception and knowledge of the agent are mapped into continuous and categorical distributions through the Probabilistic Clause Modules (PCM). These parameters are then, together with predefined laws, constraints and operator preferences, assembled as distributional clauses into a complete Hybrid Probabilistic Logic Program (HPLP). The Probabilistic Mission Landscape, obtained via inference on the resulting HPLP, is then used as basis for laying out an initial mission plan. If the proposed plan is denied clearance, a model explanations indicates critical parameters to inform an optimizer on generating a valid route. Otherwise, the planned mission can commence.
• Figure 4: Probabilistic Clearance, Explanation and Optimization results: Once a PML is constructed, the journey can be established through spaces likely to comply with the employed HPLP. Then, the clearance check validates the proposed trajectory against the PML by considering its normalized cost, which can induce rejection of the journey due to its violations of navigation constraints. Picture (a) shows such a scenario in which a proposed trajectory using a restrictive standard license at day is denied clearance. As a result, one can employ explanation (b) to get an overview of the impact of changes in the mission plan, e.g., changing to a more permissive license for special operations. Note that using a standard license at night does not entail any valid path; hence its cost is not shown in (b). While the PML seen in (c) looks similar to (a), the change in mission parameters has raised the probabilities along the optimal path and is granted clearance.
• Figure 5: Querying geospatial data from OpenStreetMap: Analogously to the HPLP in Listing \ref{['listing:uam_model']}, OpenStreetMap uses relational information that describes the types of nodes, ways and areas. Hence, we can utilize the same geospatial tags within the HPLP's spatial relations distance and over.