Oricom Technologies
www.oricomtech.com
HOME | PRODUCTS | ORDER | LINKS | ROBOTICS || PROJECTS ||

Return to:
Projects page.
See also
- NEW -
Hexapod Nico-6
"Nico-II" - Quadruped Walking Robot
Index:
|> Background
|> Movies
- (ie, walkies)
|> Design
|> Gaits

Nico-II (circa Dec 2004)
[Nico with pump-action legs]

MOBILITY
- That's what we were aiming for in the design of Nico-II. Sony's quadruped Aibo has good mobility, for a hefty $1500 - $2500 price. iCybie for under $200 can roll-over and play a few tricks, but has poor mobility. In Nico-II, rather than playing tricks or acrobatics, the goal was to build a highly-mobile walker with flexible gaits for a low price. One that could run around the house. See the movies below.

LIGHT-WEIGHT
- This is the second criterion. We've built walkers using aluminum plate and printed-circuit board for framing, and legs made out of pencils, semi-rigid plastic tubes, as well as aluminum and expanded-PVC. Gimlee-U8's 12" diameter 1/16" aluminum plate frame, although light by most standards, was still heavy for a walker. Nico-the-First's frame was lightweight pcb - strong, inexpensive, and robust. A light-weight frame requires less powerful servos, and smaller batteries can be used to yield longer runtimes. A lot of walkers use heavy machined aluminum for body parts. Pretty, but then they need stronger servos and bigger batteries. Unlike a wheeled robot, a walker has to lift and carry its own load, motors, batteries, and all, so weight is of utmost consideration.

INEXPENSIVE
- The advantage that comes along with a light frame is that cheap analog servos can be used, plus small inexpensive rechargeable batteries. Nico-II has about $80 worth of servos, and 6 NiMH AA-cells, which allows about 1-hour's worth of operation. Many biped walkers have $600 - $1400 worth of high-torque digital servos, necessary to hold the weight. High-torque digital servos use higher currents, and go through batteries faster.

Building walking robots has some of the same liabilities as being a tourist in europe or other countries. Everything you acquire, you then have to carry (or else pay the freight to ship home). Some people aim for simple and light-weight. Some people buy extra suitcases along the way :-).

BACKGROUND

"Nico-II" is an updated version of our original quadruped walker Nico. Nico-II is a further experiment in developing 4-legged walking robots, based upon vertebrate design, as described on our quadruped locomotion page.

Both Nicos have similar body plans, but original Nico had articulated (jointed) legs, with the servos for the tibias (lower leg segments) mounted directly on the upper-leg (femur) servos, while Nico-II has both servos mounted on the main body, and uses non-articulated, pump-action legs. Our octopod Gimlee-U8 also uses articulated legs, with the tibia servos mounted on the femurs.

Design Philosophy. The philosophy behind the Nico robots was to design devices which were fast and light, and could use inexpensive standard-torque servos, rather than expensive high-torque servos. Therefore, the weight of the frame was of primary consideration. Many walkers use aluminum or plastic body parts, which are beautifully-machined, but also overly heavy. This creates a vicious circle where the heavy frames require stronger servos, which require larger batteries, and lead to shorter run-times.

With the Nico designs, the main goals were: good mobility, versatile gaits, and low power consumption, meaning that smaller lighter batteries could be used. By starting from a perspective of the lightest and simplest frame possible, it makes it easier to meet the desired goals. Despite having a simple design, the use of an aluminum plate for the body on Gimlee-U8 the octopod walker added too much weight, and the device turned out to be slightly under-powered, despite having 4 legs holding up the frame at a time.

Leg Design. In a sense, the leg design of Nico-II is based upon the Sprawlita roach-like robots of Mark Cutkosky and Robert Full, the Scout-II design from McGill, and also on the leg design of horses and pronghorn antelopes. The Spawlita hexapod legs are 1-DOF pump-action legs, and the robot gets its mobility by having the legs mounted at such angles that the robot is driven forward, given proper timing between leg activations. This is a nice simple design, and the forward speed is very good, but gait flexibility is limited. In addition, the Scout-II legs are also 1-DOF, but use passive springs to give an additional "unarticulated" DOF. On the other hand, horses and antelopes (and other ungulates, such as deer) have rather long legs, with mainly bone and tendon in the lower leg segments, while muscle mass is located up close to the body. This design has a great speed and power advantage, since the legs can move very fast and the stride is long, but carries a disadvantage in that the animals are somewhat cumbersome in slow mundane movements, such as sitting down + standing up. Even though the legs have multiple joints, the relatively long lower leg segments (ie, cannon bones) carry a penalty regards certain movements. These matters are discussed on the 4-legged speed and quadruped climbing pages.

Unlike Sprawlita and Scout-II, Nico-II's leg design has 2-DOF somewhat reminiscent of horses and antelopes. Pronghorn antelopes are euphemistically said to have a "four-chopsticks-in-a-bratwurst" body type, and this is somewhat similar to Nico-II's body plan. In our Nico (the First) and Gimlee-U8 robots, the disadvantage of mounting the tibia-servos on the femurs is that this greatly increases the leg mass, and slows down leg movements. In Nico-II, however, the servos are all fixed to the main body, so the mass of each leg proper is much smaller. On the other hand, the disadvantage of the pump-action legs is that their movements are not as flexible as with fully-articulated legs, but this can be offset somewhat by building a certain amount of elasticity into the legs (something we haven't done as yet), and also by designing the legs with more degrees-of-freedom than on a robot like Sprawlita or Scout-II (which is what we did do).

An ideal leg design for a quadruped robot might have low leg mass, combined with several movable joints, but this would greatly increase the mechanical complexity of the legs. The design of Nico-II seems to be a good compromise for a simple robot. From an empirical standpoint, its mobility is quite good for walking and turning. Also, when it sits down and stands up, it does this in a manner somewhat like a deer, in figuring out how to deal with the excessively-long cannon bones in its lower-legs - see "Wobbly Deer Rising".

Capabilities Goals

  • 4-legged action similar to horses, antelopes, and other ungulates.
  • basic motions: sit-down, stand-up, walk forward + backwards, turn.
  • complex motions: traverse obstacles, search+explore, avoid getting trapped and tripped.

    Nico-II can be programmed with up to 126 different behavioral repertoires, and is being used as a testing platform for the OricomTech WMC12 Walking Machine Controller.


    <| NICO-II MOVIES

    Here are some downloadable .AVI movies of Nico-II engaged in various activities. These are low-resolution, 320x200 pixels, and the frame rate is about 4 fps, so the action is a little jerky in some cases - but the files are relatively small to download. Altogether, 15 different behaviors are shown here. All sequences appear at normal speed.

    NOTE - some people have indicated a problem with running these files on Windows machines using Media Player. However, they do seem to go ok, if first downloaded to hard-disk, and then run. In this case, instead of "left-clicking" the mouse-cursor on the filename, use "right-click" and "Save Target As". Then, run the file from the hard disk.

  • Walk Forward + Backwards - n-walk1.avi (13-sec sequence; 300 Kbytes, 15 Dec 2004). In these shots, the robot walks forward and backwards at a moderate speed of about 5"/sec (12.5 cm/sec). It can go faster on a carpeted surface. 2 separate behaviors are linked together.

  • Slow-Walk + Creep - n-slow1.avi - [in production].

  • Turn Left - Turn Right - n-turn1.avi (23-sec sequence; 340 Kbytes, 15 Dec 2004). Here, the robot turns left 360 degrees, then right 360 degrees. It takes about 8 sec to turn around completely. 2 separate behaviors are linked together.

  • Sit Down - Stand Up - n-sit1.avi (14-sec sequence; 210 Kbytes, 15 Dec 2004). Here, the robot sits down from a standing position, and then stands back up. 2 separate behaviors are linked together. The stand up behavior actually involves about 20 different timed servo movements. For this to work successfully requires that the robot first attain a stable position with 3 legs forming a stability-triangle before rising, otherwise it will flip over backwards. See the original Nico page for info on this issue.

  • Play Maneuvers - n-play1.avi (15-sec sequence; 490 Kbytes, 15 Dec 2004). Here, the robot cycles through 4 different linked behaviors, and engages in several flexible movements - called "no", "yes", "hardly", and "maybe".

  • Roam Free - n-roam1.avi (40-sec sequence; 890 Kbytes, 15 Dec 2004). Here, the robot engages in more sophisticated behavior, roaming free, and avoiding bumping into objects and getting trapped in a corner. At this point, Nico-II has only 2 sensors - namely, 2 infrared proximity-detector beams aimed 45 degrees left and right of directly ahead. In this case, there are 5 separate behaviors linked together using a real-time alarm structure - namely, walk straight ahead, plus 2 separate back up + turn sequences, which are triggered by the 2 IRPD channels. When one of the IRPDs detects an obstacle, the appropriate behavioral sequence is triggered, and the robot backs up and turns away from the obstacle. Once clear, the robot goes back to roam straight ahead. Nope - it's not radio-control. Autonomous brain.

    <| BASIC DESIGN

    Nico-II is autonomous, and carries its own battery power and locomotive devices, plus sensors and local controller.

  • body design - basic rectangle with legs mounted at the 4 corners.
  • leg movement - 2-DOF, vertical pump-action, with horizontal forward-backward movement.
  • drive - 8/ea standard 44 oz-in R/C servos for locomotion.
  • servo power - 4-5/ea on-board AA NiMH batteries (4.8 - 6 vdc, 2100 mAH).
  • gait control and real-time sensor servicing - on-board WMC12 - Walking Machine Controller.
  • high-level control - bootstrappable using subsumption techniques, or general control by a host processor.
  • size - L x W x H = 6" x 6" x 5" (15 x 15 x 13 cm).
  • weight - 27 oz (760 gm) total, including body, 8/ea servos, 6/ea AA batteries, and controller.
  • mechanical parts count - about 150.

    Sensor Array. The WMC12 controller allows up to 5 analog and 4 digital channels for real-time sensor readings and alarm generation. The following sensors are installed, or in the planning stage:

  • 2-channel IR Proximity Detector - used for obstacle detection and avoidance - mounted on the front, 2 beams approx 30 degree wide, aimed +/- 45 degree off straight ahead.
  • battery voltage monitor - for low-voltage shutdown.
  • IR Ranger - for long-range obstacle detection - [future].
  • tilt sensing using a 2-D inclinometer - [future].
  • touch sensor on each foot - we've discovered the robot can get hung-up on magazine edges, rugs with fringed edges, etc - [future].
  • roll-over sensor - [future] - Nico-II cannot right itself, if flipped over on its back (like iCybie can), but at least it should be able to shut the servos down if it flips over.

  • <| NICO-II's GAITS

    Primary Goal. Our primary goal has been concentrated towards making Nico-II highly "mobile", as opposed to performing tricks in-place, etc, as commonly seen with iCybie and Aibo. The "pump-action" legs, as described above, were specifically designed to allow Nico-II to successfully move around at a good rate of speed.

    So far the following actions have been implemented - as of Dec 2004:

  • Basic Walk - alternating diagonal walk - see here for theory.
  • Slow Walk + Creep - similar to basic walk, but slower and with greater foot overlap and better stability.
  • Trot - higher-speed form of Basic Walk - see here for theory.
  • Walk Backwards.
  • Turn Left - Turn Right.
  • Sit Down - sits on back end, like a dog sitting up - (zero-energy mode).
  • Stand Up - stands up with all leg segments pointing straight down - (zero-energy mode).
  • Lie Down - sit-down with all four legs extended forwards, similar to a deer resting - (zero-energy mode).
  • 4 Play Modes - "no" (shake in azimuth), "yes" (bob in pitch), "hardly" (bob vertically), and "maybe" (rock and roll).

    Zero-energy modes are positions which can be maintained indefinitely with "zero" energy expenditure, meaning pulse output to the servos can be shut off and the servos will not change position, as a result of external forces being in equilibrium. Actually, the servo controllers will draw a minimal amount of current, but the motors are shut down.

    Gait Implementation. This went much much faster in Nico-II than in the original Nico. This is partly learning curve, partly because we are learning more about animal locomotion every day, partly because it is easier to get good ground clearance with less complex servo movements with the new leg design, and partly because the new walking machine controller allows better interactive editing of gait sequences in real-time than with the original controller.

    In addition, once a basic gait, like the forward walk, has been implemented, it is relatively easy to change the servo movements to produce other gaits, like walking backwards and turning, etc. In some cases, certain servo movements are simply reversed in direction, while in other cases, like the trot, all it takes is to speed up the servo movements and extend the stride (ie, forward-backward range). We first implemented the basic diagonal walking gait, and it actually took only about 10 or 15 minutes to set up successfully. Of course, this was after a very lot of time spent writing the new controller firmware, and some time spent coming up with a specific mechanical leg design, and one afternoon spent building the robot itself.

    <| TOP


    © Oricom Technologies, Dec 2004