A Short bit of History
The term “robot” has been used to refer to a mechanical device that performs the same action reliably and repetitively, like a child’s wind-up toy that moves its mechanical legs and arms until its spring winds down. Similarly, mindless repetition by a human is a characteristic that is often referred to as “robotic.” Indeed, in some contexts, the term “robot” may connote some kind of humanoid or animal form that has been mechanized.
More sophisticated machines with pre-programmed features are also referred to either formally or casually as robots because they reliably repeat the same tasks. This includes most kinds of industrial robots utilized in manufacturing, including those referred to as “robot arms,” reflecting the anthropomorphic aspect of the term. Even vehicles that are solely teleoperated and remain under the direct control of their human operators, such as some Unmanned Aerial Vehicles (UAVs, or “drones”) and bomb disposal units are also referred to as robots.
But recently, the term “robot” seems to have gotten an upgrade; it has come to encompass machines and systems that are not only highly sophisticated in their functional abilities, but also are capable of acting flexibly and capably in unstructured environments. The reason for this evolution in the meaning of the term, which has been a gradual one, is a change in use-cases driven by a change in capability.
A Change in Capability
On a factory floor, stationary robot arms to do the same thing the same way, every time, because that is how quality control is maintained. Learning and adaptive software was not incorporated into these robots, as it did not enable any new functional behavior or create cost or quality benefits. Procedural execution of a defined plan has been sufficient for commercial success and there was no need to burden these robots with computationally expensive learning or adaptive software.
That is not the case with this new age of mobile ground robots. Blind procedural behaviors are no longer a sufficient autonomy blueprint. These new robots operate in complex and challenging real-world environments that require functional flexibility, advanced behaviors, and intelligence.
Only recently have these sorts of robots been commercially viable. It has taken dramatic improvements in:
- The computational efficiency of processors
- Sensitivity and fidelity of cameras and sensors
- The capability of computer vision, machine learning, and mapping algorithms
to enable autonomous and intelligent mobile robots.
Despite these enabling technological improvements, it is still astonishingly challenging to build and field mobile robots which are capable.
A Change in Make vs Buy
Why is it so difficult to build a capable robot? The primary reason is the depth and breadth of development in front of your team.
Regardless of the application or environment, your robot must:
- Gather information about itself and the surrounding world.
- Process or learn from that information to build and improve an internal model of the world.
- Determine an appropriate future goal or state and then create a plan to achieve it.
- Do all of these steps autonomously, efficiently, and safely with fixed & limited computational resources, power, weight, physical size, and material cost. And often do so outdoors with unknown and changing weather conditions.
No other kind of technological product has so many conflicting requirements. It’s no surprise that robot integrators find few compelling components to integrate into their systems, and end up building underlying technology before tackling the domain-specific efforts.
Until now. Capable Robot Components exists to change this Make vs Buy decision. We are designing a line of robot-first products which are domain-agnostic, but configurable and adaptable to the integrator’s market needs. This allows autonomous system developers to spend more time and effort on value-add, domain-specific engineering and testing.
This article was published on 2019-05-21