Structure

Structure is everywhere. An essay has a structure of introduction, main body and conclusion. A house has a structure of foundations, walls, roof. A tree, the solar system have structure. All organizations, complex entities, living bodies have structure. Within a structure, every part has its own responsibilities and connections with other parts. The clearer an entity is structured, the more functional and easier it is to improve. A body builder who wants to increase the size of his chest will go bench press, not squat. A company which wants to improve its branding will invest more in marketing, not logistics. As a roboticist, I asked myself the questions: have robots follow a structured planning (the answer is less or more). But the more challenging question is how could we bring more structure to robot planning?

Thinking in Structure

As always, it is beneficial to look at robot’s closest friend, the humankind. The rephrased questions would be: (1) do we think in structure? (2) is thinking with structure better than without it? (3) could we bring more structure to our thinking? For the first two questions, the answer is rather uncertain, as we are humans, but generally it is a “YES”. E.g., a good chess player will consider various aspects: hanging pieces, board space, king’s safety in a structure manner. Someone who is experienced in solving Sudoku puzzle will strategically scan over each number, each line, each row, each block. This also makes sense from the point of biology, specifically neuroscience. Our brains and animal’s brains are hierarchically structured. Animal brain is structured into forebrain, midbrain, hindbrain and spinal cord, in which the closer to the spinal cord, the more basic the functionalities a part is responsible for. Our brain is structured into cerebrum, thalamus, brainstem and cerebellum. Studies have shown that there is clear differentiation in task complexity which each part is responsible for. E.g.:

• The cerebrum: is responsible for high-level thinking and information processing (personality, decision-making, emotion, senses, language, etc.)
• The cerebellum: is responsible for body control, muscle memory, etc.
• The brainstem: is responsible for involuntary body functions, E.g. breathing, circulation, digestion

Structural Planning

The last and only question left is how to bring more structure into human’s thinking and robot planning? Whenever coming across ideas or approaches in other fields, I always try to find whether they have possible connections or adaptations to robotics. I came upon this idea of the “Natural Planning Model” which have various correspondences in robotics. The Natural Planning Model hypothesizes that human brain tends to follow a five-step procedure before carrying on a sequence of complex actions [1].

1. Defining Purpose and Principles
2. Visioning Outcome
3. Brain Storming
4. Organizing
5. Identifying Next Action

The arm of a small child has limitation in what it can lift. But as the child grows into an adult, the arm gains more muscle and is capable of lifting heavier objects, while still having the same shape and structure as before. I wish this framework to be the skeleton for other tools and modules, like the muscle, to attach on. If there is any current limitations (reaction time, accuracy, etc.), it should be limitations that can be tackled with advancement of other modules. The following parts cover the above steps and their corresponding robotic frameworks.

References

1. D. Allen. Getting Things Done. Piatkus Books, 2002. ISBN: 9780349408941.
2. D. McDermott. “Robot Planning”. In: AI magazine 13.2 (1992), pp. 55–55.
3. R. A. Brooks. “Symbolic error analysis and robot planning”. In: The International Journal of Robotics Research 1.4 (1982), pp. 29–78.