Biped Concept

 

This is a queued up project I hope to start after work on Quadruped4 reaches a high enough level. 

Return to the home page of Robots Australia

Robots Australia Menu

Robots Australia Home 

Quadruped4 Robot

Biped Concept Discussion

Earlier Work

Robots Australia Blog on Blogspot

My Workshop 

Servos for Robotics 

Dynamixel AX-12 

Useful Links 

Contact Robots Australia 

Biped Robot - Biped1

Concept

Biped robots, and humanoid robots in particular, always enjoy a lot of attention. Honda's ASIMO robot is generally considered to be the most advanced example to date, with the mechanical ability to walk, run (both feet actually leave the ground at the same time), climb stairs, and perform simple dexterous tasks such as pushing a trolley or carrying a tray. Probably the focus and success has been more on the hardware and the traditional control systems front rather than truly intelligent software control, with all demonstrations carefully choreographed. The power consumption of ASIMO would be quite considerable as well, with mechanical efficiency in walking far below the human gait. Whilst a tremendous (but unknown) budget has surely been thrown at ASIMO by Honda over many years of development, I do feel that there is opportunity for smaller scale projects which are perhaps more dynamic and adaptable to new technology and ways of approaching the problems to provide great steps forward, pun intended!

In recent years some very good work has been done with passive dynamic walking machines - machines that utilise gravity, and the physical dynamics of legs with passive joints to produce very impressive walking gaits with very minimal control and excellent mechanical efficiency. The biped robot 'Denise' (see Delft University Passive Dynamic Bipeds) is great to watch. 

I would like to build a machine which extends the key elements of the passive dynamic walking machine with powered joints which are able to either transition into a passive mode, or actively sense torque loading and react to minimise load in order to simulate a passive joint. A biped of this sort, combined with intelligent control could be capable of functioning in human environments with a reasonable level of efficiency. 

I propose a humanoid robot of child size (perhaps 1.2m tall), big enough to reach typical household items, created using extremely lightweight materials and using intelligent networked servo control and a small controller with radio link to conventional desktop/laptop computers running cortical simulation software.

Early thoughts are to use the Dynamixel Controller designed for the Quadruped4, and to use the Robotis Dynamixel RX-64 RS485 networked servos. These servos have a holding torque of 64kg.cm, which I may extend by gearing the output down 5:1 and using an external position feedback pot to still enable proper servo operation. This should give a working torque of 150kg.cm, and combined with a limb length of about 50cm should allow a few kilograms of weight. A very lightweight mechanical design combined with these servos which are quite lightweight in themselves should enable the biped to stand and walk - provided the right controlling software was in place!