4 Challenges Holding Back Swarm Robotics
The main factors holding back swarming robotics are the stigma of widespread robots, the lack of reliable communications, readily available distributed algorithms, and the cost of individual robots. However, these are quickly changing and the concerns can be mitigated by designing safeguards into these complex systems.
Working in the robotics space, I’ve been hearing about the robot revolution forever. It’s 2015, and instead of Jetsons-like technology taking care of the elderly, keeping our streets safe and delivering packages, all we have is the odd robot that can vacuum or mow the lawn.
Why aren’t robots pervasive? Why don’t we see groups of robots working together? Why aren’t there swarming drones for improved disaster response and unique vehicles for transportation. Many of these applications are still stuck in the lab or still in concept phase.
Here are some of the challenges slowing the adoption of these type of multi-robot swarms.
Cost of Individual Robots
Robots seem to be priced based off the traditional cost of robotic arms used in industrial automation and manufacturing. This is what the industry typically calls a robot. These types of robots can cost up to $100,000 - but they are used to manufacture cars, cellphones and other high volume, high priced products. The people who build robots are also very expensive - they are typically electrical, computer engineers and computer scientists. In some cases they have advanced degrees.
The sensors, cameras, motors and other parts that make the robot work are all expensive. And they often end up costing even more to integrate into a fully functioning robot because these parts are typically not created for robots, but are being adapted from some other type of application. One of the reasons why DJI and 3D Robotics are able to make cost-effective robots sold at around $1000 is because some of the key sensors (accelerometer, gyroscope, GPS) have become insanely cheap from the scale of mobile phones.
Robots of the past.
This hasn’t made it’s way into everything though - robot vision depends on technology like LIDAR, which is often very expensive (although some LIDAR- light types of sensors are becoming more available). As sensors and other components designed specifically for robots become cheaper, robots themselves will eventually decline in price. I go into more detail about why the costs of robots are high in a couple of previous articles here and here.
Lack of Coordinated Algorithms and Applications
As the price of robots comes down, the next major factor holding back swarming is the lack of coordinated algorithms. It is still often very hard for an individual robot to solve a problem on it’s own - and due to the cost, researchers and companies typically stick with a single robot. However, as individual robots get better and cheaper it will quickly make more sense to deploy groups of robots to tackle tasks quicker - divide and conquer.
Imagine a single robot tractor harvesting crops in a really big field - it might take a week on its own. If you add another robot and they can coordinate which side of the field to tackle, the time could potentially be cut in half. In this type of problem, coordination is fairly routine. However, it quickly becomes more difficult. Imagine groups of drones flying in formation. They need to exchange positions, direction of motion, pitch, yaw, roll, and they may need to also keep track of information about the task they are solving.
Right now drones are rarely autonomous, and at best they might have automatic gimbal control or autonomous take-off and landing. Because there is a lack of coordinated algorithms readily available, it makes it difficult to imagine what is even possible with groups of robots. Typically it is limited to a few toy applications, art shows and tech demos with few practical applications.
A single robot vision system is complex enough, can you imagine the difficulty in scaling this to multiple robots?