Complex Autonomous UAV Task Execution and Decision-Making Using s(CASP)
summary
A symbolic UAV agent called VEXERA uses s(CASP) with AirSim on Unreal Engine 5 to execute navigation, search, debris detection, precision spraying, transport and inspection by generating constraint-based commonsense plans, revising them when actions fail, and assuming perfect perception. The implemented system runs on a laptop with an Nvidia 3050 (8 GB), reports 100% accuracy in-simulator on those tasks, scales to ~10,000 objects with marginal slowdown, and has 1.2–1.3 s decision time.
AI-written summary · about
chapters (14)
- 0:00Autonomous UAVs and s(CASP)
- 1:51Why current UAVs lack complex decision-making
- 3:18Vexera and s(CASP for UAVs
- 5:24s(CASP) vs traditional ASP
- 6:46Vexera system architecture
- 9:16UAV task breakdown
- 11:02Task execution predicates
- 13:30Evaluating UAV task performance
- 15:37Experimental setup
- 17:29Performance analysis
- 18:54Ablation study on reasoning
- 20:17Scalability with object counts
- 21:43Future improvements
- 23:28Q&A hardware and implementation
related talks
19mMission-Critical AI in Space and Sky: SWaP-Constrained Intelligence Wi...
41mAutonomous vehicle development with Unity
49mMonitoring global events with AI-driven satellite swarms
57mReal-Time Decision Making in Hybrid Systems
30mCrafting a Vision-Aided Software Stack for UAV
49mUsing Deep Learning for Perception in Autonomous Systems: A Perspective from Control Theory
06mMETASAT Demonstrator: Mixed Criticality, Accelerated AI Computing for Future Space Systems
51mTeaching Autonomous Vehicles to Drive with Visible and Non-Visible Light Simulation