The Need

Electrification of high-power off-highway equipment remains challenging due to highly variable operating conditions, unpredictable duty cycles, demanding torque requirements, and limited charging infrastructure. Conventional energy management approaches often struggle to balance fuel efficiency, battery health, and machine performance when operating in real-world environments. Industry needs advanced control solutions that can adapt to changing operating conditions in real time while delivering meaningful fuel savings, reducing emissions, and supporting the transition toward more sustainable hybrid powertrain platforms.

The Technology

OSU engineers have developed an advanced, explainable reinforcement learning framework for hybrid powertrain energy management in off-highway vehicles. The technology learns how to intelligently coordinate energy sources and adapt operating decisions based on changing vehicle demands and environmental conditions. Unlike traditional control approaches that rely on predefined rules or extensive prior knowledge of operating cycles, the framework continuously identifies effective control actions while maintaining transparency into its decision-making process. The approach is designed to support robust, real-time operation across a broad range of duty cycles and operating scenarios.

Commercial Applications

• Hybrid agricultural tractors and farm machinery
• Construction and earthmoving equipment
• Mining, quarrying, and material-handling vehicles

Benefits/Advantages

• Improved fuel efficiency: Demonstrated fuel consumption reductions of up to 23% versus conventional powertrain operation
• Adaptable to real-world conditions: Maintains strong performance under uncertain, variable, and previously unseen operating scenarios
• Explainable AI-based control: Provides greater transparency and interpretability than many machine-learning-based control solutions
• Near-optimal performance: Achieves performance approaching advanced optimization-based methods while remaining suitable for real-time implementation