Photo courtesy of Iridium Communications.
HDT uses an inexpensive commercial Iridium satellite system for long-range tele-operation.
- The 20" antenna shown above does not have to be actively pointed
- Because Iridium satellites are in low Earth orbit and are at many different orbital inclinations, signal latency is low, terrain masking is generally not an issue, and the system works well at any latitude
- This antenna multiplexes together enough Iridium channels to allow real-time video
- HDT created a special low-latency H.264 video codec that uses distributed key frames to achieve excellent image quality at very low bandwidths, with very little latency
- With this system, an operator in CONUS can remotely control a Protector robot anywhere in the world
Attempts to create semi-autonomous “follow-me” robots have been unable to acquire a Safety Release from Army Test and Evaluation Command (ATEC). In trying to solve this problem, most developers add enormous cost and complexity to their autonomy kits. So far, the results have been “science fair projects” that are too expensive to ever implement as a deployable solution and are still not close to passing ATEC requirements.
- HDT took a different approach, meeting with ATEC at the beginning of our development process
- Over the course of several ATEC meetings, HDT defined Concept of Operations (ConOps) and Tactics, Techniques, and Procedures (TTP’s) that current technology could inexpensively implement and meet the requirements for a Safety Release
- As with HDT’s work in the commercial area, each part of the human-robot team does what they do best
- The human leader in “follow-me” is excellent at perception, path planning, and obstacle avoidance
- The robot’s role is to keep track of the leader, follow the leader’s path, and stop if anyone moves in between the leader and the robot. These tasks are possible within the state of the art, and they can be done both robustly and inexpensively.
- Robust performance must be multi-modal. HDT’s Protector robot does not rely on a single type of sensor, instead using a combination of vision (EO/IR), active RFID tags, LIDAR, differential GPS, Attitude Heading Reference System (AHRS), and drive sprocket odometry for each track
- Sensor modalities are all Kalman-filtered together, using rules derived from the ConOps/TTP’s agreed upon with ATEC. At least two different sensors must agree on the location of the leader – so if tall grass is blocking the LIDAR and we are in a GPS-denied area, as long as the RFID tags and the vision system’s thermal imager both agree on where the leader is, we can still follow.
- Using rugged and inexpensive components contributes to low cost
- HDT uses a planar LIDAR, rather than a more complex and fragile 3-D system
- HDT selected sensor modalities where large cost reductions are underway. Precision differential GPS systems that used to cost $40,000 are now under $10,000 and approaching $1,000 in the near future.
- Keeping tasks simple reduces the required computing power, keeping components at a lower cost and greatly reducing the challenge of cooling high-performance computers when outside temperatures are extremely high