Agile Robotics: Autonomous Forklift
The field of robotics began relatively modestly. Its founders attempted to create simple machines capable of performing tasks or interacting with the world. But in doing so, the first roboticists opened the door to an amazing area of study, rich with possibilities for extraordinary contributions to the greater good.
By exploring the burgeoning field of agile robotics, CSAIL researchers are taking those ideas to the next level. Collaboratively, the lab is trying to build “a robot that behaves predictably and naturally, and can be commanded naturally,” says Professor Seth Teller. This points to a robust, sustainable, long-term vision of robotic autonomy whose myriad possibilities suggest exciting things to come.
In the beginning, most robots were designed very narrowly for a specific task. In the case of factory-floor robots, for example, the machines were physically confined to a designated area and proscribed from human interaction. By keeping the humans strictly away from the robots, the environment in which the bot existed could be carefully controlled. But in recent years, there has been growing interest in creating robots that are able to interact with humans in a shared environment.
This need has led to the development of a new kind of robotic paradigm. To function in the spaces humans have engineered for themselves, a robot needs to be many things. Mental, physical, and situational agility, among other things, are key – the ability to evaluate and respond to a situation and the intelligence to come to a conclusion, as well as the perceptual abilities needed to sense the world around it and the physical dexterity to act according to those sensations.
The largest such project currently underway at CSAIL is an autonomous robotic forklift being developed for military application. Being developed in collaboration with Draper and Lincoln laboratories, the forklift will be one of the most sophisticated instruments the lab has ever built. It is intended to move pallet loads from place to place under voice command, and eventually to pack up an entire outdoor warehouse of pallets and transport it to a new location. Sponsored by the United States Army Logistics Innovation Agency, the forklift is designed to operate in unstructured environments such as outdoor packed earth or gravel regions.
It is unique in that, while perfectly capable of operating as an autonomous vehicle, it will also be able to function as an operator-driven machine. If a situation arose in which it encountered a task beyond its abilities, this enables the robot to be assisted by a nearby human, who could simply climb into the operator’s cabin and operate the robot through the difficulty. The challenges involved in such a project are significant, which has led to the involvement of over twenty principal investigators at the lab.
Teller is the project’s lead; other team members include CSAIL’s Stephen Proulx, Nicholas Roy, Daniela Rus, Russ Tedrake, Matt Walter, Luke Fletcher, and others from outside entities such as Lincoln Labs and Draper Laboratories. Student contributors, including Sertac Karaman, Brandon Luders, Jeong hwan Jeon, and Andrew Correa, similarly lend the project their expertise. Experts in mechanical design, electrical engineering, speech, interface issues, safety, systems architecture, control, planning, perception, and more are each bringing their particular skills to bear on the problem solving process.
Now beginning its second year, the project is arriving at a point where pallet manipulation is already a reality; the researchers are looking ahead to higher-level reasoning, improved dexterity and the manipulation of smaller objects such as individual boxes. Development of a natural command interface – designing and building a robot capable of understanding human speech, for example – is also among the many issues currently being addressed.
The forklift comes at an interesting crossroads in the history of robotics. In the past, a great deal of the code for robotics projects needed to be generated afresh for each new machine. But the spine of the forklift project is code that was written for the DARPA Grand Challenge vehicle, then lifted out and modified for this application. This represents a potentially influential precedent for the standardization of a code base in autonomous vehicles, even as the amount of code being made available online is increasing.
It is a type of collaboration that CSAIL has been pioneering since its inception. The spirit of working in teams to combine expertise is a core part of the lab that is being extended even further in this particular project. Says Professor Russ Tedrake, “I think that just in general, our robotics research is more vibrant when we’re talking more. You can talk in meetings, but solving problems together is a good way to understand the techniques that your neighbor is using and that leads to better research in general.”
This kind of melting pot of ideas acts as a catalyst, advancing the state of the art and leading to broader exploration of exciting ideas in the future. Researchers are seeking solutions to the problem of enabling robots to work alongside humans in spaces designed by and for humans, and the developments that arise from their work are likely to change the way we interact with robots at home, at work, and in the wider world to come.
Adwoa Gyimah-Brempong, CSAIL