Edge computing fundamentally changes the architecture in which applications are deployed and how resources are managed, especially in the 5G/6G era. Automated guided vehicles (AGV) are critical means of transportation in future society. We aim at investigating a privacy-preserving AGV collision-resistance paradigm leveraging edge computing. We define a model containing all potential collision occasions and propose a method to handle collisions using “virtual” traffic lights generated by edge servers. The proposed paradigm is AGV-centered, for AGVs ought not to care about where the nearby edge servers are in a broadcasting environment. In case of privacy leakage, we incorporate ciphertext-policy attribute-based encryption (CP-ABE) to protect AGV information. Besides, the use of CP-ABE leads to a location-based encryption (LBE) scheme, where AGV encrypts its running state information leveraging its current position rather than an edge server's public key. That is achievable due to the support of integer comparison in the CP-ABE access policy. To our knowledge, we are the first to implement an LBE scheme based on CP-ABE. Part of the paradigm has been deployed in a wharf yard in Shanghai. Throughout the paper, we take a yard as an example and conduct concrete experiments to highlight the feasibility and efficiency.