My impressions of Ardumower
Beautiful lawn requires constant care - watering and cutting. Obviously, the constant load on these procedures is better to transfer to robots. In this article I will talk about my impressions of the robot project Ardumower .
I must say that of the kits for self-assembly (with the ability to "poke around" in the software) - now only this ardumower is actually available. The idea to take a ready-made robot and then hack it was rejected for obvious reasons. And the experience for self-design of such a device βfrom scratchβ was completely absent, so there was no particular choice.
The kit is a chassis made of durable thick plastic in the form of a box with thick aluminum profiles that hold the walls together. Two 25-centimeter wheels made of the same plastic are rotated by two collector motors with reduction gears (up to 20-30 rpm at nominal voltage), which imposes certain restrictions on the size of the wheels - they cannot be too small driving speed will be too low. Rear wheel screwed similar to those used on movable furniture, freely rotating around a vertical axis.
The ring motor is a healthy cylinder of a collector motor weighing more than half a kilo (and its power, in fact, does not exceed 50 watts), without a gearbox, with a nominal speed of 3000-4000 RPM. On its axis it is fixed, all from the same plastic, a disk with a diameter of 190 millimeters on which three small knives are fastened, which open up under the action of centrifugal force. Due to the low power of the motor and the folding knives - it can only be mowed by the already trimmed lawn, in the overgrown area - the disc will simply be braked on the grass carpet and the knives will not mow anything. However, to maintain the lawn in good shape, with constant automatic operation - this is enough.
The design looks quite simple and reliable, although it has flaws. For example, the upper lid of the box housing, to which the rear wheel is attached, under the action of heat and load - is deformed, acquiring a small permanent deflection. Any possibility of ventilation, due to a completely closed box, is absent, which, with high ambient temperature and black color (the white plastic option was not available on the website, unfortunately) sometimes creates problems for electronics. However, there is enough space inside and you can install some kind of fan inside the case, which will provide the minimum necessary airflow for the components. Wheels without treads - tritely spoil the lawn (10-inch wheels from the trolley had to be screwed onto the running gear).
The sensor mounts are practically absent (except for holes for HC-SR04 sonars), even for such an important wire perimeter sensor, which leads to the need to independently select places and drill the case, as well as to mount the sensors on a conventional double-sided tape. For a person as inexperienced in a given area as me, this results in inevitable errors during assembly. For example, I secured the perimeter sensor amplifier directly on the front bumper, without good isolation from external conditions, which led to the oxidation of the contacts - then it was necessary to replace it and distribute the receiving coil with the amplifier, connecting them with shielded wire). The rain sensor plate would have to be fixed at an angle to prevent water from accumulating on it and the oxidation of its contacts. And so on ...
In terms of electronics - there are also certain problems. The most serious - for an inexperienced person it is difficult to understand what kind of sensors he needs for ordering. I would recommend the following set - bumper sensor, two or three sonars, accelerometer / compass GY-80 (the other may not be supported by software!), Any GPS, rain sensor, well, of course, a perimeter sensor (150 ΞΌH choke and amplifier on LM386) .
If the board assembly (the board version 1.2 available at that time was used, although a more advanced 1.3 is already proposed) is not too complicated (you need to solder and solder ...), then its scheme itself has serious limitations - for example, the width of the power conductors is insufficient for large charging currents (the charging circuit itself is not very optimal and is a linear current / voltage regulator on the LM350T), which creates problems for installing more capacious batteries or when using 12-volt instead of 24-volt components (at which current aryada increased twice). On the v1.2 board, without a lot of simple, but hemorrhoid modifications (cutting conductors, setting resistor dividers), only 5-volt arduino mega 2560 can be used (but on v1.3, translators of levels 5 <-> 3.3 promise to use more advanced arduino due).
Cheap HC-SR04 sonars are quickly clogged with dust, and they constantly cause false alarms due to the noise of the motors (the latter, however, is quite easily stopped by foam pads behind them).
The designer of the bumper sensor (air pressure sensors in plastic tubes, which are crushed when they collide with an obstacle) - generally looks like an anecdote with a mixture of SMD / SMT and normal components, with leads (the first ones are already soldered on the board, and the second ones are soldered most)
The charging station is generally invited to fully assemble both the mechanics and the electronics (you can order some of its components on the site - arduino nano, a motor driver that is used as a transmitter), which also caused many different problems.
The software for ardumower is frankly raw, although it is constantly evolving (the branch of April 26 was used). And there are two options - just ardumower and ardumower-sunray (only for arduino due with an external computer). I experienced only the first option and found a lot of flaws in it - the complete lack of GPS navigation (it is used only to estimate the speed of the robot), too simplified and constantly erroneous reversal algorithms at the perimeter (ending at this perimeter and stopping), many minor flaws (for example, by connecting a bistable relay to disconnect the battery for protection against overdischarge - I discovered that this relay does not work exactly when needed, turning off the entire robot at inappropriate moments).
I will not describe my multi-week torment with this software, I will only give a link to my modifications (emergency GPS return when it is impossible to return to its limits using standard procedures, searching for the perimeter by signal level, returning to the starting point by GPS when discharging to save time and things like that): GitHub
In conclusion, I must say that despite the many shortcomings - the ardumower project is very interesting and gives a lot of interesting impressions for all who want to join home robotics. Now the plans are to create a fully own mower, based on the experience gained, with more powerful motors (the mowing motor, for example, will be replaced with a brushless outrunner with high torque, which, although it consumes more, but surely cut off almost any grass) , such as adjusting the level of a haircut and even a solar battery (according to preliminary estimates, it will save up to 30-50% of the required energy). And, of course, the most important thing is the complete rewriting of software for the robot. I also plan to completely abandon the wired perimeter using RTK GPS navigation (using rtklib) or radio / optical tags.
I must say that of the kits for self-assembly (with the ability to "poke around" in the software) - now only this ardumower is actually available. The idea to take a ready-made robot and then hack it was rejected for obvious reasons. And the experience for self-design of such a device βfrom scratchβ was completely absent, so there was no particular choice.
The kit is a chassis made of durable thick plastic in the form of a box with thick aluminum profiles that hold the walls together. Two 25-centimeter wheels made of the same plastic are rotated by two collector motors with reduction gears (up to 20-30 rpm at nominal voltage), which imposes certain restrictions on the size of the wheels - they cannot be too small driving speed will be too low. Rear wheel screwed similar to those used on movable furniture, freely rotating around a vertical axis.
The ring motor is a healthy cylinder of a collector motor weighing more than half a kilo (and its power, in fact, does not exceed 50 watts), without a gearbox, with a nominal speed of 3000-4000 RPM. On its axis it is fixed, all from the same plastic, a disk with a diameter of 190 millimeters on which three small knives are fastened, which open up under the action of centrifugal force. Due to the low power of the motor and the folding knives - it can only be mowed by the already trimmed lawn, in the overgrown area - the disc will simply be braked on the grass carpet and the knives will not mow anything. However, to maintain the lawn in good shape, with constant automatic operation - this is enough.
The design looks quite simple and reliable, although it has flaws. For example, the upper lid of the box housing, to which the rear wheel is attached, under the action of heat and load - is deformed, acquiring a small permanent deflection. Any possibility of ventilation, due to a completely closed box, is absent, which, with high ambient temperature and black color (the white plastic option was not available on the website, unfortunately) sometimes creates problems for electronics. However, there is enough space inside and you can install some kind of fan inside the case, which will provide the minimum necessary airflow for the components. Wheels without treads - tritely spoil the lawn (10-inch wheels from the trolley had to be screwed onto the running gear).
The sensor mounts are practically absent (except for holes for HC-SR04 sonars), even for such an important wire perimeter sensor, which leads to the need to independently select places and drill the case, as well as to mount the sensors on a conventional double-sided tape. For a person as inexperienced in a given area as me, this results in inevitable errors during assembly. For example, I secured the perimeter sensor amplifier directly on the front bumper, without good isolation from external conditions, which led to the oxidation of the contacts - then it was necessary to replace it and distribute the receiving coil with the amplifier, connecting them with shielded wire). The rain sensor plate would have to be fixed at an angle to prevent water from accumulating on it and the oxidation of its contacts. And so on ...
In terms of electronics - there are also certain problems. The most serious - for an inexperienced person it is difficult to understand what kind of sensors he needs for ordering. I would recommend the following set - bumper sensor, two or three sonars, accelerometer / compass GY-80 (the other may not be supported by software!), Any GPS, rain sensor, well, of course, a perimeter sensor (150 ΞΌH choke and amplifier on LM386) .
If the board assembly (the board version 1.2 available at that time was used, although a more advanced 1.3 is already proposed) is not too complicated (you need to solder and solder ...), then its scheme itself has serious limitations - for example, the width of the power conductors is insufficient for large charging currents (the charging circuit itself is not very optimal and is a linear current / voltage regulator on the LM350T), which creates problems for installing more capacious batteries or when using 12-volt instead of 24-volt components (at which current aryada increased twice). On the v1.2 board, without a lot of simple, but hemorrhoid modifications (cutting conductors, setting resistor dividers), only 5-volt arduino mega 2560 can be used (but on v1.3, translators of levels 5 <-> 3.3 promise to use more advanced arduino due).
Cheap HC-SR04 sonars are quickly clogged with dust, and they constantly cause false alarms due to the noise of the motors (the latter, however, is quite easily stopped by foam pads behind them).
The designer of the bumper sensor (air pressure sensors in plastic tubes, which are crushed when they collide with an obstacle) - generally looks like an anecdote with a mixture of SMD / SMT and normal components, with leads (the first ones are already soldered on the board, and the second ones are soldered most)
The charging station is generally invited to fully assemble both the mechanics and the electronics (you can order some of its components on the site - arduino nano, a motor driver that is used as a transmitter), which also caused many different problems.
The software for ardumower is frankly raw, although it is constantly evolving (the branch of April 26 was used). And there are two options - just ardumower and ardumower-sunray (only for arduino due with an external computer). I experienced only the first option and found a lot of flaws in it - the complete lack of GPS navigation (it is used only to estimate the speed of the robot), too simplified and constantly erroneous reversal algorithms at the perimeter (ending at this perimeter and stopping), many minor flaws (for example, by connecting a bistable relay to disconnect the battery for protection against overdischarge - I discovered that this relay does not work exactly when needed, turning off the entire robot at inappropriate moments).
I will not describe my multi-week torment with this software, I will only give a link to my modifications (emergency GPS return when it is impossible to return to its limits using standard procedures, searching for the perimeter by signal level, returning to the starting point by GPS when discharging to save time and things like that): GitHub
In conclusion, I must say that despite the many shortcomings - the ardumower project is very interesting and gives a lot of interesting impressions for all who want to join home robotics. Now the plans are to create a fully own mower, based on the experience gained, with more powerful motors (the mowing motor, for example, will be replaced with a brushless outrunner with high torque, which, although it consumes more, but surely cut off almost any grass) , such as adjusting the level of a haircut and even a solar battery (according to preliminary estimates, it will save up to 30-50% of the required energy). And, of course, the most important thing is the complete rewriting of software for the robot. I also plan to completely abandon the wired perimeter using RTK GPS navigation (using rtklib) or radio / optical tags.
All Articles