Autonomous Multi-Purpose Defense Vehicle

The idea of enabling AI in military unmanned vehicle is not just for autonomy, but in hopes that it can do more. How much more you ask? It's safe to say that we hope to replace military personnel on the battle field with unmanned vehicles, if possible. If an unmanned vehicle can be deployed and retrieve injured personnel in the field, it would reduce the chances of injuries to medical personnel deployed into the field.

The same applies for small target search, vehicles can be fitted with a lightweight artillery weapon for self-defense. Fitted with the ability to recognize and identify the requested target. Should it be deployed into hostile territory, it may need to respond in situations if it was fired upon.

Another unmanned vehicle application is the utilization of robotics arm. The unmanned vehicle can be programmed for various purposes such as sample collection, mine sweeping, or just general road clearing tasks. The various applications bring safety to the personnel behind the controls.

Deploying computer systems for unmanned vehicle that can act as a medical retrieval vehicle requires a recognition system (cameras and sensors) that can identify injured soldiers. The soldier may be carrying a GPS beacon to direct the unmanned vehicle to his/ her site. The process of getting the injured soldier onto the unmanned vehicle may require help from another unmanned vehicle with robotic arm or other soldiers in the field. If self-defense is required when deploying into hostile area, mounting lightweight artillery may pose several challenges such as shock/ vibration from artillery fire (kickback, whiplash, etc.), confined space deployment, heat generated from the weapon, etc. Other factors that may affect the computer deploy include outdoor weather conditions (sub-zero temperatures or extreme heat), energy consumption for operation duration, etc.

The computer must also provide AI inference processing power and connectivity for sensors to detect, cameras to recognize and identify objects, actuators to control robotic arms and so forth. And in addition to the challenges mentioned, though AI can provide certain degree of decision making (recognize and identify), telecommunication capability is also a must to provide live-feed imagery for monitoring/ confirmation purposes.

Mounted with sensors and cameras, the unmanned vehicle automatically acquires images, either through object recognition or motion detection and is able to live-feed back to the control personnel for immediate analysis.

The system integrator chose to implement Neousys' rugged embedded NVIDIA Jetson solution that can thrive in vehicular environmental conditions. The vehicular unmanned vehicle is deployed into the field connected with USB/ PoE cameras, LiDARs and sensors to the Neousys' system. The use of selected cameras with IP67 waterproof characteristic, high dynamic range (>120dB HDR), auto white balance (AWB), and LED flickering mitigation (LFM) allow the system to obtain high-quality images regardless of lighting conditions, from bright sunny days to overcast weather and pitch-black nights.

With machine learning/ vision applications, it can identify targets and deploy countermeasures or actions when needed. With various certifications (EN 45545/ EN 50155, CE/ FCC, etc.) to guarantee operation in shock and vibration environments, the system’s NVIDIA Jetson SoM inference computing can operate up to 65°C ambient temperature (fanless) and up to 70°C when setup with the optional fan kit.

In addition, the internal expansion slots for wireless module installation provide real time analytic data and imagery upload, and at the same time, offer remote control capabilities to offsite personnel.