Projects
Each project starts with a real system: its signals, constraints, risks and decisions. The methods follow from that context.
Predictive maintenance
Battery and wheel degradation modelling using hybrid approaches that combine physical and neural models. Current research at LIS UMR 7020. Principal Investigator: Prof. G. Graton.
Project source ↗AI · PHM
LSTM autoencoder research on robot-battery time series. A physically informed 11-variable representation improved cycle-level Qcap estimation over the broader input set.
Environment · Sensor networks
A distributed smart-sensor concept for detecting, locating, characterising and actively monitoring forest fires while generating environmental data across complex terrain.
Risk · Geospatial
A risk analysis and assessment approach for the PACA railway network using infrastructure, maintenance, climate and geospatial data across RCP 4.5 and 8.5 scenarios.
Remote sensing · Optics
Requirements, performance and uncertainty study for passive measurement of sunlight-induced chlorophyll fluorescence in Mediterranean vegetation.
Environment · IoT
A smart-sensor concept and Fire Potential Index structure for early detection, localisation and active monitoring of forest fires across complex terrain.
Machine learning · Water
Predictive modelling work supporting a reliable estimate of the biodegradability index in industrial complex effluent.