Final Project (Fall 2015)

Simple Pick Points (max 6 points)

Every object that is successfully picked from the pick zone and placed into the deposition zone, that remains in the deposition zone throughout the trial will earn the team 1 point. The simplest version of earning these points likely involves a set of hard-coded object locations or arm configurations. You can only earn up to 6 points in this category. You may place more than 6 objects on the table in pick zone.

Timing Points

Teams will be awarded points for speed. These points will be normalized by the number of Simple Pick Points received. Ideally, teams that pick more objects will always beat teams with faster times. However, it is feasible for a team to pick very few object, but do it very fast and end up with a higher score than a team with more objects. Please don't do this.

Below is a discussion of a variety of ways that teams may add multipliers to their scores. Only the multipliers used in an individual trial count towards the trial score. If you have other ideas for multipliers, please let me know! It is up to the judges to decide if you have successfully satisfied the multiplier requirements. Make sure the judges know what multipliers you are attempting to use in a given trial.

Dynamic Part Carrier (\(m=1.25\))

A good strategy for easily picking up all 6 Simple Pick Points would be to fabricate a simple jig or part carrier that guarantees that the relative pose of all of your objects is constant. Then you only need to calibrate the pose between the carrier and Baxter, and then Baxter should be able to pick up all parts. You will be awarded this multiplier if, at the beginning of the trial, Baxter is able to automatically estimate the pose of a part carrier. In other words, I could place the carrier anywhere I want in your pick zone, and Baxter would still be able to pick up the parts.

Online Dynamic Part Carrier (\(m=1.25\))

If you are awarded this multiplier, you are also automatically awarded the Dynamic Part Carrier multiplier. The primary modification is that I should be able to arbitrarily move the carrier between picks, and Baxter will adjust, on-the-fly.

Unknown Part Locations (\(m=2.0\))

This multiplier is awarded if the translation between your pick objects can be chosen arbitrarily at the start of the trial.

Unknown Part Locations and Orientations (\(m=1.25\))

If this multiplier is awarded, you are also automatically awarded the Unknown Part Locations multiplier. The modification is that the orientation of the parts may also be selected at random at the beginning of the trial.

Dynamic Part Locations (\(m=1.25\))

If this multiplier is awarded, you are also automatically awarded the Unknown Part Locations multiplier. The modification is that Baxter should be able to adjust to translations of the parts between picks.

Dynamic Part Locations and Orientations (\(m=1.25\))

If this multiplier is awarded, you are also automatically awarded the Unknown Part Locations multiplier and the Dynamic Part Locations multiplier. The modification is that Baxter should be able to adjust to random translations and rotations of the parts between picks.

Multiple Object Types (\(m=1.5\))

This multiplier is awarded if your demo is capable of simultaneously picking two substantially different object geometries. For example, your collection of parts is made up of spheres and cubes (not different sized cubes).

Sorting Parts w/ Metadata (\(m=1.5\))

This multiplier is awarded if you are able to distinguish between two different parts and sort them into two different deposition zones with the assistance of "metadata". So, for example, you may place tags (AR, QR, etc.) on the objects to allow Baxter to easily tell which "type" of object each object is.

Sorting Parts w/ Perception (\(m=2.0\))

This is very similar to the Sorting Parts w/ Metadata multiplier, but it is more challenging because Baxter must be able to distinguish between the types of objects using only characteristics of the objects and sensors. For example, Baxter must be able to distinguish between two parts of different colors or different shapes.

Dual Arm (\(m=2.0\))

To be awarded this multiplier, you must simultaneously use both arms in a way that is more efficient than using a single arm at a time. For example, one route to accomplishing this goal would be to use one arm to capture and process images to decide the pose of the next object while simultaneously planning and executing a pick-and-place motion with the other arm. You can not get this multiplier by simply toggling which arm you use to pick and place objects.

User Interface (\(m=2.0\))

You can receive this multiplier by implementing a user interface that allows an operator to control the operation of the picking and placing operations. For example, you could provide an interface that allows a user to select which object should be picked or placed next. Another example would be providing an interface that allows a user to correct mistakes in perception. For example, let's say you try and estimate the orientation of an object on the table, and your image processing algorithm misjudges the orientation by 30 degrees. The user interface could allow the user to correct this angle to improve grasp reliability. Making a meaningful and useful interface is likely a bit tricky, but it is also likely a good strategy for increasing the level of robustness that you are able to achieve. Note that creating an nice, high-level interface that allows a user to completely direct Baxter (with no on-board perception or autonomy) may disqualify you from other multipliers. Correcting a misjudged orientation is fine; completely specifying the orientation with an interface will disqualify the orientation-specific multipliers from above.

The judges have several categories, where they can assign bonus points if they choose. The categories are roughly defined below, but ultimately, it is up to the judges' discretion and interpretation of the categories to decide how and if to allocate these points.

Ambition (max 5 points)

Points awarded for attempting a particularly ambitious task or solution strategy.

Creativitiy (max 5 points)

Points for solutions to common problems that are unique or creative.

Technical Implementation (max 5 points)

Points for uniquely solid technical implementations of solutions. This could include techniques that work very robustly, or strategies that require higher levels of technical competence to implement.

The overall score will be calculated with the following equation:

\[ S_t = \left( N + \frac{N}{T + P}\right) M_{tot} + J \]

where

• \(S_t\) ~ The total score received for a trial. This is the score that will be used for ranking teams.
• \(N\) ~ The number of Simple Pick Points received (max. of 6)
• \(T\) ~ The total time of the trial being scored (expressed in minutes).
• \(P\) ~ The penalty time for interventions (expressed in minutes). This time is calculated by multiplying the number of interventions in the trial (max. of 3) by 2 minutes.
• \(M_{tot}\) ~ This is the total multiplier received for the team during the execution of the trial. This is calculated by adding up all multipliers that were employed during trial execution (as determined by the judges).
• \(J\) ~ The total points awarded by the judges.