Hello, this is the IIT Bombay Mars Rover Team,
We just wanted some clarification regarding the on-site computing infrastructure for Challenge 1.
In the Technical Handbook, the On-Site Computing section states:
There will be 5 physical machines… Each Team will be assigned a single virtual machine with a fair share of the host’s resources (CPU, RAM, disk and networking). … We do not expect any high-performance GPU access to be available during the Competition.
Meanwhile, the Husarion Panther manual for the CAM06 – Stereolabs ZED X warns:
ZED series cameras require NVIDIA GPUs to be fully operational. Image acquisition can be performed on the CPU, but some functionalities… are not available. We recommend pairing ZED series cameras with PC07 or PC09.
Our Concerns & Questions:
- How do you plan to support four ZED X cameras—each ideally requiring an NVIDIA GPU—using only the Intel Iris Xe integrated graphics on the ASUS NUC or on-site VM?
- Since real-time depth mapping and other GPU-accelerated SDK functions are critical for autonomous tasks, will there be any provision to offload or stream processing to an external NVIDIA GPU (e.g., via on-site networked GPU host), or must all computation truly run locally on Iris Xe in real time?
Any clarification on these points will help us design our perception pipeline correctly. Thank you!