This is a general purpose packer designed for apples but capable of packing other fruit. The prototype has packed apples in Nelson. It is the first of eight systems to be installed. Each unit can pack 150,000 apples in a 24 hour period or 18 million apples in an apple-packing season of about 120 days. The system accepts bulk apples, singulates them and then rotates each apple until its best side is uppermost while the axis of the apple is horizontal. It then places the apple with its axis aligned in the tray. It uses two proprietary robots and a sophisticated vision system to accomplish this.
Play video(Please note that this may take several minutes to load)
This is an autonomous 4-wheel drive robotic vehicle which performs the following functions:
Future Applications
In existing kiwifruit packhouses, approximately 30% of the fruit is rejected on the basis of size and quality. The fruit growers pay the packhouse a packing fee which is based on the gross tonnage with a fine for rejects. The vision software on the automated picker will be developed to recognise fruit which is undersize, unripe, misshapen or marked. Consequently, more of the fruit going to the packhouse will actually be packed for sale.
Pollination is an expensive and difficult operation in kiwifruit orchards and unexplained hive deaths are a considerable worry to orchardists. Consequently, some orchardists apply pollen manually so that they are not reliant on bees. Manual applications do not apply the pollen efficiently. The vision system on the automated kiwifruit robot will be developed to recognise female flowers and apply pollen precisely to the flower in an optimal manner (leaving sufficient room between pollinated flowers for the fruit to develop in an unobstructed way) using a customised pollen delivery system attached to the robot hand.
The pruning of kiwifruit vines is another expensive and time-consuming operation for the industry. The autonomous robotic system will be adapted to perform this function.
The robotic arms of the system will be adapted to pick other types of fruit such as apples and oranges.
Image
Junior is a humanoid robot with the following features:
The stereotypical manual wheelchair has many disadvantages; it is inefficient, cannot surmount the smallest object, requires both hands and is hard on the hands. It has very limited mobility for any user who is not strong and energetic. Essentially, it offers no gearing and although wheelchair ramps are becoming more prevalent, many people do not have the strength to use them.
Based on these limitations, we developed an alpha prototype of a manually operated four wheel drive wheelchair which, in its crude first attempt, offered the following features:
The beta prototype is currently being developed and has:
An intelligent vision system has been developed which is capable of recognizing the features of objects, even when they are partially obscured. Further details are available in:
Bakker, H. H. and Flemmer, R. C., "Data Mining for Generalised Object Recognition"
Howarth, J. W., Bakker, H. H. and Flemmer, R. C., "Feature-based Object Recognition"
• Uses GPS and intelligent vision to navigate kiwifruit orchards; manoeuvring around obstacles such as posts and recognising braces at the end of each row.
• Identifies fruit, discriminating for size and gross defects. Picks the fruit and places it gently into the bin. Checks the fruit level at each point in the bin and adjusts fruit placement to fill the bin evenly.
• Decides when the bin is full, goes to the end of the row and unloads the bin. Searches for and picks up an empty bin with its forks, returns to the last position and resumes picking.
• Operates 24-7, checks for light level and operates floodlights if necessary. Checks for rain or dew and covers the bin with a tarpaulin when this is detected so that picked fruit is protected.
• Goes into secure mode (for example when the fruit is wet), moving the robotic arms to a safe position, switching the unnecessary power systems off, and maintaining power only to the main (monitoring) computer and radio link. Wakes up when appropriate and resumes picking.
• Receives and responds to communications via radio link and uses voice recognition to respond to verbal commands.
• Uses a variety of recovery strategies to deal with faults such as getting stuck, vision becoming obscured, etc.
• Collects data on the fruit yield from a particular orchard.
• Sophisticated machine vision which allows him to learn and recognize objects and to estimate their position and orientation
• Artificial intelligence which allows him to identify objects (nouns), get their attributes (adjectives), observe their movements in time (verbs), characterize their movements (adverbs), observe their spatial relationships (prepositions) and see whether a pattern of behaviour always follows on another pattern (logic)
• Two biomimetic arms and hands, a head with cameras for eyes, mobility, radio link communication with a master computer
• Machine Personality: curious, cautious, tireless
Image
1. Single handed operation – the user could carry a cup of coffee while going from one place to another.
2. Surmounting obstacles such as those shown in the accompanying images.
3. Travelling at 20 mph on black top when operated by a person of average fitness.
4. Reclining comfortable seat whose design is medically superior to the traditional hammock seat of most current wheelchairs.
5. Travels comfortably over grassy or gravel surfaces.
6. Exercises the full upper body rather than primarily the triceps (as in the case of the manual wheelchair).
7. Offers gearing so that very weak people can climb wheelchair ramps, albeit slowly.
8. Little friction associated with the hand grip for propulsion.
• Frictionless adjustable-gain drive for each arm
• Push-pull operation
• 20:1 selectable drive ratio
• Steering by either hand using servo system
• Hydraulic disk brakes
• Self-charging battery for steering, radio, lights, flashers
Image
Flemmer, R. C. and Bakker, H. H., "Generalised Object Recognition"