Meet SCOUT: The basics of our robot’s software and hardware

Here on our blog, we have talked about using robots to prevent wildfires, removing organic waste from forest floors, and even creating the software that makes this possible. However, it can be easy to overlook how all this fits together. How do we actually plan on doing this? While the details are complicated, the technical basics of our robot come down to three main parts: the robot’s body, its processing system, and its code. This week we met up with graduate researchers Kevin Chau and Dhawal Modi to help us understand the various parts of our robot.

The robot’s “body” is called SCOUT–specifically, SCOUT version 2.0–and is manufactured by AgileX. It is an unmanned ground vehicle with four wheels, motors, and a battery with 150 minutes of runtime [1]. "SCOUT is very powerful," says Kevin. "It’s meant to go over rough terrain and pull heavy loads." Dhawal elaborates: "SCOUT is suited for the kind of environment our team plans to use it in: the forest." It can travel at 1.5 meters per second (~3.3 mph) and haul an impressive 50 kilograms (about 110 lbs) [1], which will be important down the line when our robot needs to carry wood waste. SCOUT is the machinery itself, the part of the robot that moves. However, in order to move, it needs to have something tell it when and where to go.

[1]

Figure 1

AgileX SCOUT 2.0 Model [1]

Mounted on the top of SCOUT is the Jetson, a computer stationed inside a plexiglass case. This small integrated computer–also known as an embedded system–controls the SCOUT robot. Kevin explains, “The robot itself can’t ‘think’, so it needs a ‘brain’ to do all the logic and reasoning.” The Jetson computing platform itself is made by NVIDIA and is officially called an AGX Jetson Orin module [2]. The Jetson includes a powerful integrated central processing unit (CPU) and graphics processing unit (GPU). This operating system is specifically designed for robotics and power-constrained edge AI applications. While the CPU is needed for most of the Jetson's inner-workings, the GPU is designed to help the CPU perform parallel calculations like matrix calculations, explains Dhawal. "GPUs are primarily used to render graphics,” Kevin adds. “For example, a PlayStation or XBox needs one to function. These graphics calculations are similar to ones needed for a lot of AI algorithms.” We need the high-performance, parallel computing capabilities of the advanced GPU within the Jetson, because many of the AI calculations deal with large streams of data such as those that come from the camera input images used for SCOUT’s computer vision software. The CPU alone cannot do this efficiently, and the GPU by itself cannot carry out the baseline functions needed to run a computer that are handled by the CPU. Hence, the CPU and GPU need to be tightly integrated together, along with specialized software for coordinating both, in order to have a working AI application.

The final and third part of our robot is the code. Our engineers write and input code for the Jetson that allows SCOUT to navigate and maneuver. The code is then conveyed to the SCOUT’s robotic motors and steering system, so that the Jetson can control the SCOUT and make logical decisions, such as following a target or avoiding an obstacle in the road. Our developers write code for the Jetson in Python and C++ programming languages. They utilize Robot Operating System (ROS), which is a free, open source software framework used to write programs for robotic applications [4]. Dhawal described ROS as having a steep learning curve, but once you are past this curve it is easy to implement and flexible in terms of integrating different functionalities, such as the ability to follow a person. If you would like to learn more about our software, check out this blogpost.

[2]

[2]

Figure 2

NVIDIA Jetson AGX Orin Developer Kit [2]

To recap, the three main parts of the Robotic-Assisted Wood Waste Removal Project's actual robot includes: the SCOUT robot car body, the NVIDIA Jetson AGX Orin computing platform, and the application software (ROS framework). According to Kevin, the key is trying to understand how these components fit together. Together these components will allow us to make a robot that can clean up organic waste in forests to prevent dangerous wildfires [3]. Through finding the right fit for these three main parts, we can make our communities and forests safer. Additionally, Kevin mentions, "Our solution could also be adapted for a wide range of applications beyond wildfire cleanup, such as (self driving vehicles, personal assistance robots, elderly health care, package shipping and delivery, etc.)"

Special thanks to Kevin Chau and Dhawal Modi for walking us through all the technical information used in this post!

Sources:

1. “Scout 2.0.” Agilex Robotics, Agilex, global.agilex.ai/products/scout-2-0. Accessed 7 Aug. 2024.

2. “Nvidia Jetson AGX Orin.” NVIDIA, NVIDIA, www.nvidia.com/en-us/autonomous-machines/embedded-systems/jetson-orin/#:~:text=Jetson%20AGX%20Orin%20Series,%2C32GB%2C%20and%20Industrial%20versions.

Accessed 7 Aug. 2024.

3. Pereira, P., Bogunovic, I., Zhao, W., & Barcelo, D. (2021). Short-term effect of wildfires and prescribed fires on ecosystem services. Current Opinion in Environmental Science & Health, 22, 100266. https://doi.org/10.1016/j.coesh.2021.100266

4. “ROS - Robot Operating System.” ROS, ROS, www.ros.org/. Accessed 13 Aug. 2024.