Robotics Engineers -- How could a bot navigate by temperature?

[grumpymrgruff]

I've seen a bit of research that relies on imaging methods like FLIR for input. My question adds a couple wrinkles:

How could a robot

• Navigate without imaging? I.e., using only local sensors like a thermocouple or array of thermocouples.
• Extract navigation info from turbulent thermally heterogeneous media?

I'm thinking about some research showing that lobsters and other marine animals may use their antennae to read fine changes in water temp, but I'm at a loss to come up with how they might extract usable info without sitting still to average temperature differences across antennae and estimate the thermal gradient. By all appearances, they don't do this.

 

[Grinkle]

Without making assumptions regarding the environment, I don't think navigation by thermal gradient is a well posed problem. Thermal gradients are caused by many things, and the navigation question needs to be bounded by a context if it is to be solved - I don't think there is a general answer.

A species that evolved underwater near the edge of a steam vent might evolve a particular algorithm for processing thermal data, and a species that evolved in an underwater environment at thermal equilibrium except for the occasional biologic might have evolved a completely different algorithm for processing thermal data.

Edit :

I am not a biologist. As an engineer, I will note that thermal data is very low bandwidth, and there would need to be a compelling reason for me to design something that relied on a low bandwidth information source if for the same cost I could design something that could make use of a higher bandwidth source. Infrared radiation that is emitted by a warm object is higher bandwidth, for instance, than thermal data per se. So if I were making a lobster, either thermal sensing antenna need to be really really cheap compared to chemical sensors or EM sensors of some frequency, or there needs to be no other data of higher bandwidth available for me to use.

 

[anorlunda]

How can we answer your question without knowing your requirements? Is the robot navigating the aisles of a china shop, doing brain surgery, or trying to locate an island in the ocean? Must the navigation path be accurate to 1 cm or 1 km? Are the temperature differences between objects 1 degree or 1000 degrees? How fast does the robot move? What is the required angular field of view? (or just straight ahead?)

The more you share your objectives and requirements with us, the better the quality of the answers.

 

[donpacino]

I've seen a bit of research that relies on imaging methods like FLIR for input. My question adds a couple wrinkles:

How could a robot

• Navigate without imaging? I.e., using only local sensors like a thermocouple or array of thermocouples.
• Extract navigation info from turbulent thermally heterogeneous media?

I'm thinking about some research showing that lobsters and other marine animals may use their antennae to read fine changes in water temp, but I'm at a loss to come up with how they might extract usable info without sitting still to average temperature differences across antennae and estimate the thermal gradient. By all appearances, they don't do this.

like grinkle, i am an engineer not a biologist. I want to point out lobsters have eyes. It is likely they use both the antennas and eyes and touch (possibly) to navigate.

To implement a system like that, one approach is to write an algorithm around a kalman filter to merge the data between the two sensors. the filter would also help remove noise if designed correctly.

unfortunately my answer is general because your problem was general.

 

[dragoneyes001]

wouldn't the antenna be more for self preservation? if they do sense temp changes it may only be to avoid entering overly warm water areas which would make them not part of the navigation and more like a cats whiskers to inform that a path is safe to continue on.