How could we build self adaptive robots (easy to configure and re-purpose) more effectively? In this article, we argue that modularity and Artificial Intelligence (AI) techniques are key. And we explain why.

The figure depicts three different strategies for building robots: a) the traditional approach, b) a modular approach where interoperable modules can be used seamlessly to extend the robot and c) the Modular And Self-Adaptable (MASA) approach where modules, besides interoperating, are detected and configured automatically in the robot for its use.

Source and extended article: “Towards self adaptable robots: from programming to training machines" Víctor Mayoral, Risto Kojcev, Nora Etxezarreta, Alejandro Hernandez and Irati Zamalloa

The most popular process in industry to build robots (a) leads to machines that lack of flexibility and reconfigurability.

Any modification in the robot during what we call the “critical section” will demand a re-execution of all the following steps. It is a time and resource-consuming approach for building robots.

In the modular approach, b), the integration effort is removed and the so called critical section reduced significantly. However, the task of programming robots remains cumbersome. New modules are interoperating but need to be introduced in the logic of the system manually. This implies that for each module addition or modification, a complete review of the logic that governs the behavior of such robot will need to happen. In other words, the adaptation capabilities of these systems are still limited.

Section c) illustrates the Modular And Self-Adaptable (MASA) approach for building robots that radically changes the robot building process.This way, rather than being programed, modular robots train themselves for a pre-defined task. By continuously integrating the information from its modules, based on an information model like HRIM, the robot is able to adapt automatically when new modules are added. This approach reduces both the human development effort and time significantly.


The MASA strategy (c) for building robots can be summarized as follows:

  1. Buy module: This step refers to the action of acquiring those robot modules required to build the robot. Since the devices are modules, they are assumed to be interoperable, easy to integrate and re-use.
  2. Define task (critical section): This step refers to the process of defining the goal that the robot should accomplish in a mathematical form so that the learning algorithms can be based on it. Typically, in Reinforcement Learning (RL) –a set of AI techniques– this mathematical expression is captured in what is called a ‘reward function’ that steers the learning process of the robot.
  3. Robot assembly (critical section): We capture the physical construction of the robot in this step, which is simplified since all modules interoperate.
  4. Automatic training (critical section): In this step we implement AI techniques that allow the robot to continuously integrate the information from its modules and adapt dynamically a neuromorphic model to fit the task defined in previous steps. This way, regardless of the physical changes that happen in the robot (such as additions or removal of modules), the robot will automatically retrain itself for the task.
  5. Deploy: Once trained, this approach outputs a flag that notifies about the success or failure of the automatic training step. In the case of failure, the user can refine the task definition (step 2) or add additional modules to the robot (step 3) and allow the training process to iterate again automatically (step 4) until success.

In the article, we present an experiment that aims to shed some light into the relevance of this new approach for building robots meant for real-world scenarios (subject to noise and errors) and on the spot testing. We compare the results obtained by the traditional approach (a) and the MASA one (c) on a given task: to reach a given point in the workspace.

The setup consists of a robot with 3 Degrees-of-Freedom (DoF) in a SCARA configuration. The robot is built and configured by following the traditional (a) and MASA (c) approaches. The configuration of each robot follows from its building process and is either programmed (a) or trained (c).

Simulation is used to accelerate the process of experimentation applying, when appropriate, faster than realtime techniques.

Results show that the new approach proposed for building robots outperforms the traditional one in the presence of noise by even an order of magnitude in some cases.

The figure pictures arbitrary trajectories of each one of the outputs of MASA for the given task under different levels of noise. The values in the x and y-axes are in meters (m). An interesting observation is that in the presence of Gaussian noise, MASA is able to adapt, and although the trajectory overshoots the target, it returns to the goal improving its accuracy. For example, the yellow line presents the output of MASA for a robot where each joint has been subject to an error of zero mean and 0.1 radians (5.73 degrees, refer to Table 1 for more details) of standard deviation. Still, it manages to get to less than 2 centimeters.

You can find the detailed experimental set up and results in the original paper: “Towards self adaptable robots: from programming to training machines".

Interested in knowing more about our research on AI? Find our most recent research papers here.