Last month, we wrote about autonomous quadrotors from the University of Pennsylvania that use just a VGA camera and an IMU to navigate together in swarms. Without relying on external localization or GPS, quadrotors like these have much more potential to be real-world useful, since they can operate without expensive and complex infrastructure, even indoors.
One potential application for drones like these is disaster operations, but honestly, that’s just what everyone says when you ask them how their mobile robot could potentially be useful. What’s much more interesting to us are commercial applications, and with drones, that inevitably means talking about delivery. There are a lot of reasons why we’re skeptical about most commercial delivery drones, but that doesn’t mean that the idea of using drones to move things from place to place isn’t a good one.
Vijay Kumar’s lab at UPenn has been working on using their GPS-independent quadrotors for transporting payloads, and they’re doing it collaboratively—the idea is that objects that are too large or heavy for one quadrotor to move can instead be moved by multiple quadrotors working together, and ultimately, they could be the best way to move items around a warehouse.
The use of multiple MAVs can provide additional benefits compared to the use of a single vehicle when solving a task. Although the complexity of such systems increases with the number of vehicles, the additional vehicles allow the transport of payloads that cannot be transported by a single vehicle because of size and payload constraints, and can provide robustness to the system to compensate for single vehicle failures.
What’s new here is not the transportation of objects with multiple quadrotors but rather doing it without some kind of external localization system. Each of these quads is using its own VGA camera and an IMU, and that’s it, meaning that what you see here would work just as well outside, or in your living room. You can read lots more about how this works in our previous article.
While each quadrotor can do a decent job estimating its position from camera and inertial data alone, that estimation will gradually drift away from the drone’s true location, getting worse and worse the farther the drone moves. With two (or more) quadrotors rigidly connected while transporting an object, you can combine the location estimates from each robot to optimize both of their positions, resulting in a much more accurate estimate that drifts less.
Better localization means better, more reliable performance, and even with cargo, the video above shows the quadrotors zipping around at speeds of 4.2 meters per second and accelerations of 5 m/s2, a “level of agility and autonomy [that] has never before been accomplished at this scale,” according to the researchers.
For more detail, as well as a glimpse at a future of drone warehouses, we spoke with lead author Giuseppe Loianno.
IEEE Spectrum: Can you summarize how this work relates to the work we reported on earlier?
Giuseppe Loianno: This work considers multiple aerial platforms cooperating to transport a payload. In our previous work, we considered the coordination among robots to perform tasks that are inherently distributed. A more advanced form of collective behavior is required for tasks that simply cannot be accomplished by individuals but can be accomplished by cooperation. Examples of cooperation are seen in cooperative manipulation and prey retrieval in nature. In cooperative manipulation, each robot needs to interact with the payload (and therefore the environment) and also accommodate the rigid constraints introduced between the different robots.
What are some of the unique challenges in controlling quadrotors that are collaboratively transporting an object?
The challenges in cooperative control required a new approach which allows independent control of each vehicle while guaranteeing the system’s stability. The estimation, planning and control algorithms are designed for this new “system." For example, the localization subsystems are independent for each vehicle, and thus may provide different estimates of the load position and orientation. And this will in turn result in control actions that are not consistent with the rigid body constraint.
Would this approach scale up to more quadrotors collaborating to transport larger, heavier payloads?
The approach we proposed scales very well with large number of robots. From a localization point of view, we also show in the work that the optimization solution does not depend on the number of the vehicles carrying the structure. The control and planning approaches are also independent.
You mention in the paper that transportation tasks in warehouses is one potential application. Can you describe how such a system might operate?
Imagine you are in a warehouse where there are objects that are either too heavy either too big to be transported by a single vehicle. These objects need to be identified, picked up and moved to a final destination. In the future, we aim to have a complete system that will be able to automatically infer to an operator the number and types of vehicles needed to pick each object in a coordinated fashion and transport them to the final destination. Moreover, the algorithm will allow multiple teams to concurrently pick different objects in the warehouse guaranteeing obstacle avoidance and solving the overall transportation task in an optimal and distributed way.
What are you working on next?
We are working on pursuing experiments with multiple quadrotors and studying the tradeoffs between increased control authority and increase in inertia of the system, between improved localization estimates because of cooperation and the increase in complexity resulting from an increase in the number of constraints. We are also interested to estimate the geometry and inertia of the payload during the task. Finally, we are interested in combining this work with our previous work on grasping to perform a richer repertoire of tasks.