Edge-Cloud AI for Dynamic Pricing in Automotive Aftermarkets: A Federated Reinforcement Learning Approach for Multi-Tier Ecosystems

Bhuram, Shiva Kumar (2025) Edge-Cloud AI for Dynamic Pricing in Automotive Aftermarkets: A Federated Reinforcement Learning Approach for Multi-Tier Ecosystems. World Journal of Advanced Engineering Technology and Sciences, 15 (3). pp. 126-135. ISSN 2582-8266

Edge-Cloud AI for Dynamic Pricing in Automotive Aftermarkets presents a novel federated reinforcement learning framework that addresses the unique challenges of pricing optimization in fragmented supply chains. The architecture enables collaborative intelligence across manufacturers, distributors, and retailers without centralizing sensitive data, preserving privacy through differential privacy guarantees while maintaining high pricing accuracy. A modified multi-agent deep deterministic policy gradient algorithm reduces training variance compared to standard federated approaches, while a cloud-based meta-optimizer resolves cross-tier supply-demand mismatches. The system was evaluated in both simulated environments with thousands of SKUs and real industry deployments, demonstrating faster convergence than centralized reinforcement learning despite data fragmentation, robust regulatory compliance, and significant profit margin improvements. The architecture's selective parameter update mechanism substantially reduces cloud computing costs and communication overhead while maintaining model performance, establishing a new standard for privacy-preserving collaborative intelligence in multi-tier retail ecosystems.