RISC: Resource-Constrained Urban Sensing Task Scheduling Based on Commercial Fleets
summary
RISC implements an MDP-based scheduling system that predicts vehicle locations, builds a vehicle–region graph, and casts each time slot as a set-cover selection of a subset of vehicles (actions are subsets, state tracks remaining sensing hours per vehicle) to avoid duplicate uploads and save cellular cost. Implemented on Shenzhen fleets with 14k taxis, 13k buses, 4k trucks and evaluated with an 11k personal-car trace for a tracking app, achieving about 10% average coverage improvement over greedy baselines.
AI-written summary · about
related talks
21mRAS: A Resource Allowance System for Perpetual Region-wide Resource Allocation
40mCo-operative robots sharing the load
55mScaling out operational research with Rust
06mD2Park: Diversified Demand-aware On-street Parking Guidance
07mUrban Map Inference by Pervasive Vehicular Sensing Systems with Complementary Mobility
20mRoute Optimization Engine for the Real World
17mApplication of Online Resource Allocation Based on Ray
22m"How We Built a Rust-y Real-Time Public Transport Map"