Vice President of Hardware Engineering and Robotic Manipulation
Distribution yards are inhospitable, oftentimes hazardous places. Manually-driven yard trucks repeatedly move trailers to be loaded, unloaded, cleaned, and stored. At Outrider, we’ve set out to autonomously move these trailers around the yard. We developed a self-driving yard truck, but we also had to figure out how to automate the task of connecting the trucks to the trailers.
Today, yard truck drivers get in and out of their cabs as many as 6 billion times per year to connect pressurized brake lines to trailers. This releases the parking brake so the truck can move the trailers around the yard. It’s a tedious and oftentimes dangerous job that must be completed around heavy-duty equipment in all weather conditions. Truck drivers must use their fine motor skills with one or both hands to connect the truck to the trailer by way of the gladhand.
Thus begins the tale of TrailerConnect and where my expertise in robotic manipulation has enabled me to lead a team of exceptional engineers to solve this challenge. My background in robotic manipulation ranges from wearable exoskeletons and advanced prosthetic devices for the disabled to autonomous ground vehicles with robotic arms to navigate and manipulate complex environments, such as in improvised explosive device defeat or search and rescue operations.
Robotic manipulation is a challenging field because many manual tasks we humans consider simple – say, taking the lid off a jar or using scissors – are quite complicated for robots to complete, but come very naturally to humans.
When it comes to autonomously moving trailers, I put this experience to work. Outrider needed a robotic arm that could perform a human action – connecting the brake lines. To complicate matters, the connection points on the trailers called “gladhands” vary dramatically.
There are tens of thousands of combinations of gladhand types, locations, and mounts with varying degrees of freedom – something a human knows how to deal with almost intuitively, but a robot would need to see, learn, and be cleverly constructed and equipped to handle. By the way, the alternative – modifying all those trailers circulating in the supply chain – was out of the question. Semi-trucks connect brake lines to trailers by way of “gladhands” (shown above). TrailerConnect robotically connects to tens of thousands of gladhand types, locations, and mounts combinations.
So, we decided to build a robotic manipulation system called “TrailerConnect” that automates connecting the brake line to the trailer using deep learning, a machine learning technique that can help robotic manipulation systems more effectively accomplish what comes naturally to humans.
Like other robotic solutions I’ve worked on, TrailerConnect moves and makes decisions to accomplish the task of brake line connections in the harsh environments of distribution yards. As each task takes place, TrailerConnect senses, navigates, and interacts with its environment – through unique ways of combining sensors, software algorithms, and electromechanical systems.
TrailerConnect connects truck air lines to trailers so they can be autonomously moved around the yard.
Based on my past experience, I knew that designing a robotic arm to mimic human movement was not going to be a viable approach to connect trucks to trailers. It’s one thing to get a robot prototype up and running in a pristine laboratory, benchtop setting, or well-defined environment or application – it’s another beast entirely to create one that works with precision in rain, shine, or snow in harsh industrial settings. That’s why we assembled a stellar team with experience in a multitude of fields.
Inside the TrailerConnect design and development team.
This diversity of experience enabled us to develop hardware and software at a fast pace, test the arm in operationally relevant environments, and use data to inform development steps to optimize the system. To create TrailerConnect we combined proprietary software algorithms, hardware, and sensors that we integrated on top of a base commercial off-the-shelf robot arm. This vastly extends the capabilities of that base manipulator to become the system we need for our application.
All of this is to say, what’s easy for us as humans isn’t necessarily easy at first for a robot. The human sensorimotor system that primarily includes the eyes, regions of the brain, arm, and hand, is arguably the most capable tool in the natural world. But with the right programming, training, and design, robots are capable of extremely precise and repeatable motions that can eventually approach and eventually surpass human speed and performance in some applications.
We are building a system to accomplish this with TrailerConnect. After more than four years of rigorous design and testing, we have created a system that can accomplish this task, thereby freeing up yard workers to focus on less grueling activities. As our technology continues to mature, we look forward to continuing to provide robust, reliable robotics solutions for autonomous yard operations that will surpass human capability.
Matt Johannes, Vice President of Hardware Engineering and Robotic Manipulation
Matt Johannes is the Vice President of Hardware Engineering and Robotic Manipulation at Outrider. An experienced robotics and unmanned systems research and development engineer, he’s passionate about creating novel technical solutions in a variety of challenging problem domains for use in real world applications. Previously serving as Supervisor of Robotic Systems at the John Hopkins University Applied Physics Laboratory, Johannes has robust and deep experience leading and managing large multi-disciplinary development teams.