1 m hexapod (autonomous insect)


Skip to comments
robot video thumbnail

Hi guys,
this is the first introducing part of a series, describing an autonomously operating hexapod with about 1 m width (my large insect).
Idea behind this project: From early childhood on I wanted to build a kind of mechanical turtle or similar, being able to survive in the garden from spring to fall and doing whatever it would be programmed for, recharging its battery, seeking for shelter in case of rain, etc.
Decades later (meanwhile a retired physicist, phd after 30+ yrs. in the space business) it was really time to realise this dream, meanwhile based on modern state-of-the-art electronics and software.
Then, a colleague suggested to acquire an EZ-B V4-2 as the core element for such an undertaking and from thereon things picked up speed quite rapidly.
As mentioned above, goal is the realisation of a quite large autonomous hexapod.
The figure below is illustrating the general shape of this robot as well as the nomenclature for the legs (1, 2, , 6) and the servos (1.1, 1.2, , 6.3), respectively.

User-inserted image

User-inserted image

The base plate is made from 10 mm plywood and the legs from 20 mm x 20 mm pine wood. servo base blocks: 20 mm pine wood.
A further light weighted plate (2nd floor) will be added later for accommodation of a scanning lidar, etc..
The hexapod’s legs are moved by 18 servos (Longrunner B07K68MK3Y:

User-inserted image

In order to suppress unwanted rapid servo motions during EZ-B initialising, a switchable power distributor was added for each side (servos 1.1 through 3.3 and servos 4.1 through 6.3, respectively). These units are served by a central 2S2P LiPo battery (10.000 mAh) and are protected by individual fuses and are galvanically connected to the EZ-B.

User-inserted image

Only the signal pin of each servo is directly connected to the EZ-B. servo Vcc and GND are both fed by these power distributors.
Result: During EZ-B initialisation all servos are switched off - to be activated via toggle switches as required. Of course, all servos need to be positioned adequately beforehand.
Currently, the hexapod is looking like this (together with his silent friend).

User-inserted image

User-inserted image

Programming was done in Synthiam BLOCKLY; surprisingly, the final version was completed within two days only !

Upcoming activities: First steps in the garden (Forward direction).
Then: Backward, Turns, Sideways.

So long, that’s it for today,



By — Last update
#46   — Edited
Lots of options for speech synthesis. There are several text to speech solutions and there are also speech settings (male female) but you may want to poke around in the skills section to find a voice / language  you like. There is Bing text to speech and Watson as well as others.

Project -  add - audio
#47   — Edited
Hi guys,

the 1 m insect has received some major modifications:

Previous power distributors incl. switches, fuses and sockets for the servos have been replaced by a kind of fuse banks, each equipped with an individual fuse for each servo.

Background: During a recent test, one of the servos ran into a hard stop (unnoticed). After a minute or so, Amperes were smelled, but it was too late, the servo was already dead (overheated electronics). The 2 common fuses (serving 9 servos per side) did not react, since their values had been defined to withstand currents for 9 servos under load and therefore could no protect a single unit in a hard stop. As a result a new concept with individual fuses.

User-inserted image

Previous version (1 common fuse).

User-inserted image

New version (1 individual fuse per servo plus common switch) I am aware of this one missing fuse;) .

This concept (one unit per side) works very well and will surely save nerves:D (Fuses: Cheap and Servos: Expensive) !

Work on the upper deck was also progressing as planned.

The receiver for the digital 8-channel remote control as well as the second EZ-B have been installed on the upper deck.

User-inserted image

User-inserted image

Upper deck with receiver of the remote control and the second EZ-B.    OK, cables still need to be arranged. Tentacles (here at the left) not attached; only their microswitches visible.

Latter unit will control all present and future sensors; avoiding signal lines running from the upper deck down to the lower deck (other EZ-B; now only responsible for the drive servos and some housekeeping stuff). Just a single power line is running from the upper deck down to the 2S2P5000 LiPo battery, making future disassembling and inspection much easier.

The insect is currently equipped with the following sensor suite:
-    Three ultrasonics (middle deck), covering 180 of the front section.
-    Two tentacles (upper deck), acting as final obstacle warning at zero distance.
-    Camera (upper deck), function still t.b.d.   .

For future long range outdoor navigation (primarily for autonomous location of the battery charging station) an optical solution is currently under test.

Due to the fact that homing-in towards e.g. the charging station has to work during bright sunshine as well as during dusk or dawn.

Classic optical concepts, based on lamps or LEDs in the VIS or NIR spectral range will hardly work, sincethe SNR at the detector outputs will be too low, even when using extreme narrowband optical filters <-- damned expensive !

VIS = Visible section of the electromagnetic spectrum (380 nm to 750 nm).

NIR = Near infrared section of the electromagnetic spectrum (800 nm to 2500 nm).

A way out could be, using Mid-IR signals (wavelength about 10 m). This is the regime of thermal signatures.

Mid-IR = Middle infrared section of the electromagnetic spectrum (3 m to 50 m).

Important: Keep in mind that the sun is not emitting any thermal radiation or heat. Maximum intensity of solar radiation is at about 500 nm (green); therefore leaves, etc. are green for optimised plants’ metabolism !

Heat in sunlight is exclusively created on / in illuminated surfaces (conversion of solar radiation with short wavelength (VIS and NIR) into thermal radiation with longer wavelength (Mid-IR).

The following diagram (spectral intensities of sun light (VIS and NIR) and Mid-IR radiation) is illustration this effect:

User-inserted image

Planck curves for solar signal (T = 5778 K) left and terrestrial Mid-IR radiation (T = 373 K) right; both well decoupled.
Abszissa: Wavelenght in metres. Ordinate: Not scaled.

The sun is about invisible in the Mid-IR !

As a matter of fact, garden environment in the Mid-IR is pretty dark, only the small warming effect of sunlight on leaves, grass or other objects may generate thermal signatures of lower intensity or contrast.

Principle of Mid-IR outdoor navigation:
An object with a hot surface (about 100C), aka Black Body in the centre of the charging station is acting as an optical transmitter and heat sensitive detectors on the insect are delivering information for the navigation towards this station. Due to its high SNR relative to the environment, this thermally bright object will easily be detected by suitable sensors.

Realising such a hot surface is not a big deal. It was more interesting, whether suitable detectors were available.

Answer: YESSS ! !

PIR sensors, typically in use for activating the light at the entrance of a house are commonly available at low cost (small versions including electronics below 2 Euros).

Typical example: HC-SR501.

User-inserted image

HC-SR501: Compact PIR sensor for Mid-IR outdoor navigation.
Board size: 28 mm x 32 mm.

The upper deck will receive six of these sensors; easily covering the 360 surrounding. Their digital outputs will directly be fed to the upper EZ-B.

Dedicated subroutines in the main control software will transform their information into corresponding signals for the motion servos.

More details about this hopefully working - concept in the near future during / after outdoor tests.

Well, that’s it for today !


Wow that is so neat to see progress on this! Specially the wave length filters. I’m excited to witness the succsss!
Hi DJ,

sorry to say, but there are no filters around.

It is just mother nature that cares for the separation of these spectral ranges (VIS + NIR vs. Mid-IR).

In fact it is one of the advantages of such an approach that no filters at all are required.
#50   — Edited
Hi guys,

the large insect is about receiving its heat-sensitive sensors for later locating the charging station for its battery, once strolling around in the garden.

A hot black body source (T appr. 100C) will be installed at this station (size of a flat iron) so that the hexapod will be able to home in from any point; just picking up the heading to this target and start following this thermal signal.

As mentioned in a previous contribution, HC-SR501 boards, comprising a small PIR sensor element, will be used. These are normally delivered with a wide-angle lens that needs to be removed for the current usage.

User-inserted image

HC-SR501 board with PIR sensor wide-field lens removed

Six of these sensors mounted at the upper deck - are covering the complete 360 azimuth around the insect.

Unfortunately, operation of these PIR sensors is jointed with a major disadvantage: They only react on changes of thermal signatures, meaning that they are only generating a single HI output pulse if a warm object is entering or leaving their field of view (fov), respectively. Pulse duration is adjustable within a few seconds.

If such a warm object would quietly stay within their fov, no further HI pulses would be generated (no intrinsic re-triggering).

Required: Modulated thermal signal.

Potential solutions:
-    Modulation of the warm transmitter (black body) itself is not possible due to its high thermal inertia.
-    Modulation of the thermal signal at the receiver is quite easy: A small cardboard pendulum with a central aperture is mounted immediately in front
     of the sensor aperture.

Due to the hexapod’s rumbling motions, this pendulum will swing; periodically covering the sensor aperture.

User-inserted image

Sensor in housing with pendulum in front (sensor yet not well centred)

By this neat mechanical device, the received black body’s thermal signal is amplitude-modulated: ON-OFF-ON-OFF-ON-OFF .., meaning that the HC-SR501 sensor is periodically re-triggered; thus generating a continuous HI signal as long as the black body will stay within its fov.

Basic requirement: Pendulum periodicity needs to be shorter than a single sensor HI pulse.

The following figure is illustrating the principle of this pendulum (here simulated by tilting the sensor housings):

User-inserted image

Left: Pendulum at maximum left deflection (sensor blocked),
Center: Pendulum at centre (sensor free),
Right: Pendulum at maximum right deflection (sensor blocked).

Problem solved !

The little sensor units with their pendulums are mechanically sensitive and do require sturdy bodies (figure below). A white coat will ensure that their interior will stay cool (thermal sensors inside !), even during hexapod operation in bright sunlight.

User-inserted image

Six sturdy sensor bodies (white coat still to be applied)

In order to minimise payload mass, these bodies as well as the insect's mid deck and upper deck had been made from 10 mm sandwich foam plates (both sides covered with 0,2 mm aluminum.

Next steps need to concentrate on suitable Blockly software routines so that the hexapod will reliably pick up the heading towards the black body (charging station) to commence homing in towards the charging station.

So, that’s it for today


Wow!! I’m thoroughly impressed. Can’t wait to see it in action! Thanks for sharing the details. I find your build interesting
Hi DJ,

do you see a change that camera commands in Blockly could also be indexed ?

The "large insect" is currently undergoing a kind of "metamorphosis" towards an autonomous instrument carrier.

I am currently thinking about an onboard digging device in the rear for autonomous capturing of soil samples that had been identified as "interesting" by the front camera.

A second camera (mounted in the rear) would be very useful for monitoring these sampling activities.

Thanks for a brief comment !


Hi DJ,

just installed ARC 2021.02.02.00 on another PC (Panasonic Toughbook) and noticed that an important BLOCKLY feature is missing:

The file managment line "Load Workspace      Save Worspace",

normally located between the lines "Create Variable" and "Camera Preview".

OS of this device is WIN10 Pro.

On my other PCs (desktops and laptops), everything is OK.

Any idea, why I cannot handle BLOCKLY files on that machine ?

Many thanks in advance !


Ensure you're viewing the Blockly interface and here are the load and save workspace buttons..

User-inserted image

Unless by chance you are viewing scripts of a robot skill, in which case there is no "workspace" in robot skills.. Robot skills may be imported using the Merge option which allows selecting skills from other projects to merge.

User-inserted image

of course I had selected the correct window; I am working with Blockly for many months now.

My fault: I had forgotten to add a figure, showing what I meant.

here is this Blockly window without load and save workspace buttons:

User-inserted image

It seems that also the camera preview window in the lower left is not being loaded (compared with your example).

I will now remove ARC via Windows system tool and will re-install it; maybe this will help.
After ARC re-installation: 50 % success  -  this PC (Panasonic Toughbook) is operated by touchscreen and pen.

Hitting the potential location for one of the BLOCKLY save or load buttons with the pen, one of them will appear (see below).

User-inserted image

User-inserted image

A bit strange behaviour, but this is OK for me, since these functions are now available.

However, the section with the Camera Preview (normally immediately below these save and load buttons) is still missing
#57   — Edited
I dont have that problem. Must be that laptop of yours...
Try other resolutions perhaps?
Have no problems with all my other desktops or laptops; it' s most probably related to this particular machine ...

... will dig deeper over the WE.
Now that is the strangest behavior I’ve seen. I cannot think of how that can even happen. There’s nothing different about those buttons and layout than any other buttons in arc. 

It is a good idea to use a recent version of the ARC software. Difficult to diagnose with an older version. 

make sure the tablet has the latest windows updates as well.
I use ARC 2021-02-02-00.

The Toughbook has a WIN 10 Pro version from summer 2020 (younger than all my other PCs).

But I will check for its WIN updates.
Same result upon installation of WIN updates.

Must be the machine itself.

No tragic, since these buttons can be brought to life via tipping the pen.
#62   — Edited
Hi guys,

here the latest and final images for the large hexapod (formally the "large insect"):

User-inserted image

User-inserted image

User-inserted image

AND at least virtually ...

... it made it to MarsxD   !

User-inserted image

Background: JPL NASA

Due to other projects that need to be finished within March, further field testing of the hexapod's autonomous  outdoor navigation, etc. have been shifted to early April.


Hi DJ,

the hexapod is currently equipped with 2 (two) EZ-B.

How should the second one be connected to the PC ?

As client as the first one ?

The EZ-B manual is not covering operation of more than one unit.
Looks good at a first glance - THX !