Testbed for Semantic Compression-Based Networked Immersion Using Boston Dynamics Spot

This work introduces a practical implementation of the cXR+ semantic compression method within a real-world test environment. Utilizing a Boston Dynamics’ Spot Dog robot, cXR+ is deployed over a network where the robot semantically compresses data from its physical surroundings. It uses color-codes (CCs) to convey the essential information needed to recreate an immersive environment at the destination. Creating an immersive environment typically involves five phases, each producing a specific output: calibration, registration, volume reconstruction, marching cubes, and rendering. In the cXR+ approach, the volume reconstruction phase is replaced by a client-side virtual network function (VNF) that compresses the registered point cloud into CCs. These CCs are transmitted over the network to a server-side VNF, which decodes them to generate voxel frames, completing the volume reconstruction. We evaluate the cXR+ method against baseline approaches such as raw data transmission, JPG/PNG compression, and compute-and-send techniques. The results demonstrate that cXR+ achieves the highest compression ratio of 98% and the lowest aggregate compression and transmission latency of 0.83 seconds, validating the findings presented in [1].