Data-driven on-line load monitoring in garbage trucks

The payload of garbage trucks may vary substantially over the time, affecting both the vehicle performance and driving safety. Information on the load in real-time could thus play a key role for monitoring and diagnostics. Unfortunately, physical sensors directly measuring the vehicle mass are usually too costly for commercial trucks, while the correlation between consecutive values of the load is not considered by most of existing approaches for mass estimation. Since this correlation characterizes load variations in garbage trucks, this paper proposes an ad-hoc approach for payload estimation, which relies on inertial sensors only. To minimize the tuning effort, we introduce a strategy to automatically select the key tunable parameters of the estimator. The effectiveness of the proposed approach is demonstrated on experimental data collected on a real truck.