Meet: Max Beutelspacher
Max has been with NODE since the early start at Fraunhofer IPA and works as a Team Leader for the Scheduling and Planning Chapter.
During our NODE T-Talk, we talked about understanding fleet management systems, strategies for adapting to changing schedules, ensuring flexibility and safety in execution, and productivity gains through effective planning.
Hi Max, tell us a little bit about yourself. What did you study, and how did you get into the world of robotics?
I originally studied physics and during my studies, I had absolutely no contact with robotics. Instead, I focused on numerical simulations in the field of cosmology. It wasn't until I did an internship at Fraunhofer, in the working group that later became NODE, that I got into Autonomous Mobile Robots (AMRs). I found the field so fascinating and full of opportunities for further development that I decided to start working there after completing my studies, focusing on controlling and planning robot paths. After almost 3 years on the team, it was an easy decision to switch to NODE.
Which aspects of this domain have particularly sparked your interest?
I have always been fascinated by the applications of basic mathematical algorithms, which, however, either do not fully solve the problem or do not do so quickly enough and therefore need to be adapted to the specific problem. In addition to the pure theoretical efficiency of the algorithms, implementation details often play a significant role, such as caching behavior or maintainability, determining whether the result can be successfully used for years. Ultimately, I am always excited to see when individual puzzle pieces come together to form a big picture, that is when many individual components come together to create a functioning overall product.
Can you give us an insight into your daily workflow at NODE Robotics and the tasks you undertake in the coordination, planning, and scheduling department?
I lead the Planning Team, which means that I am often the point of contact if a customer's robot does not behave as expected. Since robot behavior depends on many external factors, it is sometimes not easy to understand from the outside. This part always gives me a good feeling for the areas where we can further develop our current planning stack. Furthermore, I coordinate the development roadmap, create implementation proposals for internal or external feature requests, and try to anticipate potential problems. And finally, I also work on the actual development tickets daily, testing them in simulation and on the robots, and exchanging ideas with my team members about the specific implementations.
How would you explain to someone who has no idea about the subject what a Fleet Management System is and how it works?
Let's imagine we have a bustling marketplace full of people doing their shopping at the weekly market. Everyone is in a hurry, trying to reach their destination as quickly as possible. This works well enough if there aren't too many people and there's enough space, otherwise, everyone gets slowed down. It's like a fleet of robots; as the fleet size grows, there's a need for global coordination so that the robots don't necessarily take the shortest path but the fastest one. Communication among the robots about where they're heading is crucial, much like how humans communicate by eye contact when they meet and can avoid each other early on. Additionally, you could optimize the order of shopping; perhaps you go to the cheese stand first, where the line is short, before buying vegetables. (Ultimately, you realize early enough when you're hungry and grab a snack). All these tasks involved in coordinating larger robot fleets in a confined space and optimizing task distribution and execution by gathering information from all parties involved are handled by the Fleet Management System.
In dynamic environments, changing plans is inevitable. What strategies does your team employ to adapt to changing plans and ensure smooth execution of tasks for robot fleets?
The robots continuously monitor their surroundings using laser scanners or cameras, allowing us to adjust the path if obstacles collide with the robot's path. If the path is completely blocked, we first try to find a new plan in the immediate vicinity, and if that fails, we search for a new path across the entire map. Ideally, this happens in the background before the robot stops. In the past, we have also experimented with methods where robots inform each other about blocked paths, allowing them to replan before encountering the obstacle. However, in practice, this has not yet proven necessary.
How do you ensure flexibility in execution, particularly concerning quick responses to local obstacles and preemptive route changes for safety?
Initially, the robots prioritize safety, meaning the safety control triggers if the scanner's safety field is breached, which enlarges at higher speeds and in curves. This means the robot always stops in normal operation before getting close to an obstacle. However, we aim to avoid triggering this safety system as it brings the robot to a halt. Therefore, we model safety fields in our software and predictively check them against the sensor data we receive, incorporating them as input into path planning optimization. This sometimes entails taking a slightly longer detour around obstacles if it allows for faster travel there. Another aspect is ensuring the predictability of the robot's behavior for operators and staff, so we strive to maintain the original plan for as long as possible. Additionally, with a zoning model, we offer the option to instruct the robot strongly on which paths to prioritize.
Can you provide some examples of how effective planning and coordination have led to significant productivity increases for clients?
We operate in many environments where numerous other robots, conveyor trains, or forklifts are active. Even as a human, navigating through these areas can be challenging at times. The robot must constantly navigate dynamic obstacles there. If we strictly followed our tracks there, everything would come to a halt. Another example is warehouses where another client operates their robots. These warehouses consist of narrow aisles often blocked by pulled-out crates at different heights. As soon as we perceive this, we must immediately find a new global plan, as the robot (so far) unfortunately cannot tidy up.
What challenges do you see for robotics' future, particularly in intelligent coordination and efficient planning?
The two major challenges I see specifically in planning are the ever-increasing density of robots and other autonomous systems, which can only be addressed through better global cooperation, as well as more complex kinematics and safety controls as increasingly small conveyor trains or forklifts operate autonomously. It's also a significant challenge to convey the complexity of a robotic system to humans without overwhelming them. While I have many parameters to adjust the robot's behavior, I do not want to burden the user with that complexity. They should be able to understand and achieve the behavior they desire with just a few clicks and options. Hence, our goal is to make Robots easy to use!
_
The interview was led by Nadine Trommeshauser.
