In a fast-changing world where automating human work was already emerging, COVID-19 further accelerates change and automation due to social distancing and it creates economic challenges by interrupting everyday life. Working with more distance from each other will be the new normal and machines will have to work together with humans for them to keep them safe, but also keep them in business.
We use our autonomous technology to support, or sometimes take over, the dirty, dull and dangerous work of humans. That is why we develop robust autonomous systems for reliable operations in industrial applications. In this article we want to highlight technical challenges we had to overcome and the exciting possibilities of these new technologies.
With AGV’s being developed in the 1970’s, autonomous mobile robots have been around for quite some time. Due to their high costs however, their usage has been limited to special applications . With the advent of the open-source Robot Operating System (ROS) and cost down of electronics the total cost of ownership for an autonomous mobile platform decreased dramatically. But as with all disruptive innovations, there are still some challenges that the open source ROS stack does not tackle.
ROS in its core is a middleware framework, which allows different software components to communicate not only within a single robot, but also across multiple machines. On top of that, the open-source community has developed a stack of software modules offering a wide range of functionalities for robot localization, navigation and data visualization, among others. Despite this, developing robust robotic applications that comply with industry expectations remains a challenge. Most of ROS modules have been mainly tested in lab prototypes, without focusing on robustness, reliability and other type of requirements that are specific to industrial applications. Therefore, at Nobleo we have been building ROS robotic solutions complying with most industrial standards.
One of the most challenging tasks of a robot is to localize itself in its environment. A common method used by the open-source community is the so-called Advance Monte Carlo Localization (AMCL). For most industrial robotic applications AMCL does not offer a suitable solution due to poor accuracy and drift. To tackle these issues, at Nobleo we make use of a sensor fusion module that processes data from different sensors like Lidars, Inertial Measurement Units (IMUs), RTK-GPS, and dedicated optical sensors like the optical odometer Nobleo has helped develop for Accerion.
For robot navigation we have developed and verified our own stack to offer high quality solutions to our customers. For instance, we have partnered with the ROS-Industrial initiative to develop a Full Coverage Path Planner (FCPP) and a highly customizable path tracking PID control algorithm, both of which will be open sourced in the second half of this year. Additionally, we have developed a path tracker that is especially suitable for big and heavy mobile robots like automated tractors and forklift trucks. Here we also offer a custom Model Predictive Control (MPC) algorithm to dynamically adjust the vehicle velocity when needed.
Caption: For robust software that functions over several robot platforms, a well-defined architecture is very important. The interfaces between the modules need to be the same for all different robot-configurations. This ensures reliable automated software testing and it makes it possible to keep all software releases stable and mature.
Our latest and most exiting projects are unfortunately customer-confidential for the time being, but we have proven to the need for our navigation and localization solutions in several past projects. To illustrate where autonomous mobility can help in commercial processes, below some examples for the dirty, dangerous and dull automations.
A good example is the WasteShark autonomous waste-collecting boat we developed for start-up RanMarine. Because the vessel is underactuated (no sideways thrusters), strong currents or wind will affect the following of straight lines. This robot uses RTK-GPS, Lidar and accelerometers to determine its position, direction and speed. The localization solution is plotted on the known map and compared to the wind direction and given planning. This ensures that all the trash is collected. The Lidar is also used to detect unexpected obstacles in the water, which will cause the plan to be paused until the way is clear. Because a PID controller is used, standstill at this point is possible; the WasteShark will act as if it were on a virtual anchor keeping its front at a fixed point, compensating for winds and currents.
Our localization solution is flexible enough that it can also accommodate sensor solutions for indoor localization as is required for warehousing solutions. Our warehouse AGV solution can bin-pick items and deliver them to the packing belt. Opposite to other fields of navigation flexibility is not the most important feature in industrial applications. Robustness and predictability are. Sticking to a generated path is therefore part of the AGV navigation solution.
The tank cleaning robot developed for the oil-and-gas and chemical industry cleans industrial tanks with 3000 bar water pressure by driving in horizontal lanes over a vertical wall by using magnets in-between the wheels. But the soiled wall and the gravity pull will result in a significant lateral slip. Also here, the PID controller used in our navigation solution covers these disturbances and follows the line prescribed by the full-coverage planner.
Caption: Clockwise, starting upper left: 1. Tank-cleaning robot, using 3000 bars of water pressure to clean the inside of oil tanks or strip the paint of the outside of the tank. 2. Warehouse item-picking robot; autonomously driving to the correct location and picking the prescribed product from the shelve. 3. WasteShark autonomous waste collection boat; using wind-direction and GPS to collect floating trash from water ways.
There is a growing demand for automating the dirty, dull and dangerous work since many industries are already short staffed and the importance of a safe and healthy work environment is being seen in more and more places. This will not be a threat to jobs but an opportunity for technology and humans working together and be more efficient and most important: to be safe.
A key feature we have developed for interaction with our robots and between the robots themselves is a shared database ‘world-model’ in the cloud. Here, the user can install the robots and configure their tasks and the robots can use it to align their actions amongst themselves.
Nobleo Technology has been active in this field since 2016 and we have been developing mobile robots with an all-star team that gained experience with e.g. building the TU/e Robocup robots, Fontys picking challenge robots and several robotic start-ups. Today a core-team of over 10 people strong is working on robots of all kinds.
The Nobleo Autonomous Solutions team is the biggest asset in ensuring high-quality code, bringing decades of experience to the table.
By investing in an identical architecture for all robots we can quickly gain traction in new projects. While the actual set of executables and libraries may vary, all robots have the same building blocks and interfaces. This leads to stable and robust software components, even when work is being done on multiple robots at the same time.
Nobleo Technology has delivered a dozen autonomous robot designs to date. During the development of these robots the team identified where the open-source ROS software is valuable, but also where improvement is needed for real-world applications. So, if you have a challenge that you think could be solved with an autonomous solution, we will be happy to help.