Hi all,

I'm currently having my semester break before going continuing for my last semester for my degree in Mechatronics Engineering.

So.

I've given myself a deadline, that is by the end of this break, I should have decided on what my future path will be.

1. Continues postgrad and my ultimate aim, to be a researcher/lecturer.
2. Grad and work as an engineer, hopefully in an R&D department.

My area of interests are currently .... not quite as refined.

they include.

1. Robotics
2. AI


I like electronics/programming.. and I like to learn more on AI.. machine learning and etc.

But..

This decision will take me possibly 3 -4 years down the road.

So.

I hope to know more about a postgrad's lifestlye. financial opportunities, and responsibilities.

Thank you for chipping in your opinions. =D

If your supervisor has a research grant, you can work as his research assistant.

May I know what is your final year project?

There are many types of modern robots i.e. Robotic arm, Mobile Robot (crawler), Mobile Robot (wheeled), Humanoid robot (Android), Unmanned Aerial Vehicle (UAV), Autonomous Underwater Vehicle (AUV), Swarm robotics. What kind of robots that interests you?

If becoming a researcher/lecturer in Robotics & AI is really your ultimate aim, I can recommend your CV to a researcher in controlling UAV with Smart Material Actuator using Shape Memory Alloy. Of course, you must have a sound basic knowledge in Control Engineering and proficient in MATLAB Control System Toolbox, which I believe you have already learned the foundations in Mechatronics. Below is a video of a flapping-wing UAV.



Machine learning usually concerns the study of robotic systems that can learn from sets of data. Could you describe what kind of algorithms (e.g. Neural Networks, Genetic Algorithm, etc.) that you intend to use in your final year project?

My fyp is in the development of a Bio-Inspired Swarm of Mobile Robots that is able to(hopefully) Localize itself and the target.

basically, my project milestones would be

1. Develop self localization ability, with the limitation that the environment is known before hand, by using beacons.
2. Extend the ability for the swarm to be able to know where it is currently.
3. After which it will have a search strategy for a target as a swarm, that it will converge to.
4. That will ultimately give us a rough map/path towards the goal.



The practical applications of such swarms would be seen in disaster scenarios, where rescuers need to search for victims under rubble, as an example. It takes the idea of Swarm Intelligence from nature, such as Ants, Bees, and apply those concepts of co-operation between agents. The idea of a swarm of low-cost agents that has distributed processing means that no single agent is crucial to the mission, which means, throw one bot? it still works. Second bot fails? it still works.


But... currently, I'm still in the 1st phase, where I've overlooked the ... details needed in developing self-localization ability. It's all rosy in simulation and concept, but hardware implementation/algorithm? Nightmare. haha. Localization's a vast area of research on it's on... SLAM is an ongoing research.. even in the military.
Still persevering though. Might consider some lost in accuracy to trade for project completion.


Honestly, I'm a person that have zero interest in mechanics. I'm more of a code/concept guy.
Also my control theory is not exactly sound. I can use it, but I'm not fond of it.

All my semester projects till now have concentrated on electronics and software.
Cheap Pulse Rate Detector that clamps onto ur fingers.
Cheap energy saving controller circuit for automatic lamp posts.(abit lame, but the concept and circuit is sound)
Vision Based Intruder Recognition
and etc..

I'm learning on my own about neural nets and genetic algorithms(GA). Planning to implement it in my fyp, when my fyp's underlying foundations are set.



Any comments?

I have found my way. TQ all. =)....

Hi Ivanov,

How's the future of your baby Bio-Robots? Start crawling?

=)

well, in simulation at least


they are able to have an emergent behavior of path chaining just from simple rules.

imagine releasing a bunch of these micro swarms in future to search for victims in disaster areas. you'll be able to follow the paths created by them to the victim to save them.

now i'm working on getting position information from the path itself, so that before you even follow the path, you already know the coordinates of the victim at the end of the path.

after your fyp, have you considered quantitative finance since programming(c++ i hope) and modeling is your interest? you could try to google quantitative analytics/ financial engineering

I am currently in midst of indecision. haha.

am in the process of weighing in all pros and cons and future.

I am currently in midst of indecision. haha.

am in the process of weighing in all pros and cons and future.

MicroBots = "Ants"
Victim = "Food"

Ant colony optimization (ACO) is based on the cooperative behavior of real ant colonies, which are able to find the shortest path from their nest to a food source. The idea behind ACO is based on pheromone deposit and pheromone evaporation. An ant travels from node to node until it reaches the food node. The first ant finds the food source, via any way, then returns to the nest, leaving behind a trail pheromone. Before returning to the nest node (backward node), the last ant deposits some pheromone on arcs it has visited.

Because of the increase in the pheromone, the probability of this arc being selected by the forthcoming ants will increase. After each search process, trails of pheromones evaporate, leaving only the strongest pheromone associated with good or promising paths. In short, the more pheromone deposited on the global best path, and the better the exploitation ability.



Conditions:
---------------
(1) The location of the victim is known.
(2) The environment is known or at least partially known.

Homing Beacons
----------------------
One idea is to use IR-emitting homing beacons, i.e. the equivalent of a lighthouse. With two IR beacons with different frequencies, the robot can determine its position at the intersection of the lines from the beacons at the measured angle. The advantage of this method is that the robot can determine its position and orientation. However, in order to do so, the robot has either to perform a 360° rotation, or to possess an omni-directional vision system that allows it to determine the angle of a recognized IR beacon.



As can be seen in Figure (top), knowing only two beacon angles is not sufficient for localization. If the robot in addition knows its global orientation, for example by using an on-board compass, localization is possible (middle). When using three light beacons, localization is also possible without additional orientation knowledge (bottom). You can substitute the light beacon with Infrared (IR) beacon.