Difference: GeneralizedBiped (1 vs. 11)

Contribution: Propose a new control strategy for physically simulated walking motions that is simple and generalizable. Each of the 4 components that composes it is clearly explained, justified, and its relevance tested.

Evaluation:It seems like they tested their system in a variety of scenarios and while performing different tasks or after various bodily deformations. But, as is usually the case, they do not provided any comparison to real data (motion capture in particular) neither visual nor numerical.

Reproducibility: An open source implementation of their project is provided, which is basically the best one can hope for in such cases. However, they clarify that this is "one implementation" and that they use various parameters "in their results".

Improvements: The authors acknowledge themselves that their characters cannot move at high speeds (i.e. fast motions make the characters fall over). There is no comparison to real data, nor does there seem to be a real measurement of robustness.

-- Main.ginestra - 25 Nov 2011

Contribution: The authors provide an integrated strategy for walking controllers that combines a number of techniques used in virtual control strategy over the past few years. Specifically, they describe 4 components:

• Motion Generator : Joint angles on a character are specified over half a walk cycle using splines (then mirrored for a complete cycle); at runtime, PD-controllers seek to track the specified angles.
• Foot Placement : Robust balance and some degree of velocity control is achieved by foot placement planned using an inverse-pendulum model.
• Velocity Tuner : Using a virtual force derived from the difference between the desired and actual velocity of the character and the Jacobian of the center of mass, a torque is determined that is applied when possible at the stance ankle to move toward the desired velocity.
• Gravity Compensation : Similarly, the Jacobian of the center of mass of each link is used to apply torques that negate the effect of gravity, -mg.

Evaluation: The paper describes experiments in which the controller successfully handles variations in gait parameters, like walking speed, and character proportions and morphology. They also demonstrate the generality of their controller by highlighting several tasks such as picking up objects or carrying crates that are easily incorporated into the controller.

Reproducibility: The authors provide high-level descriptions of each of their system components, but do not describe specific parameters or details. However, the authors do provide source code and executables from the project web page, so the results are indeed testable and reproducible.

Improvements: In general, the paper's writing style is clear and logical. The specific contributions, methodology, and limitations of the control system are laid out in the introduction, which helps provide a roadmap for the rest of the paper. One evalution that may have been helpful is a quantitative comparison of the "naturalness" of their gaits compared to motion capture data, as in "Optimizing Walking Controllers". Do the different gaits of varied characters reflect real-world differences correctly (at least for the physically possible characters)?

-- BenHumberston - 25 Nov 2011

Contribution This work allows designing diverse walking styles using more complex controllers such as inverse kinematics, viretual forces and trajectories.

Evaluation: The variety of tests under different circumstances shows the robustness of the solution.

Reproducibility: It is reproducible

Improvements: There are some details about the implementation that are important and not included in the paper.

--Ernesto Torres

-- Main.etorresv - 24 Nov 2011

(a) Contribution: This paper combines several techniques to create a more robust/natural biped walking control strategy. There are four key components: PD controller for target trajectoires, gravity compensation, balance-aware foot placement and continuous balance adjustment by mimicking the effect of a virtual desired force. The strategy supports are large range of motions including picking up objects and climbing stairs.

(b) Evaluation: Most of the results are qualitative. This includes sequences of images generated by the simulation or descriptions of the behaviour. This is to be expected because we're measuring robustness and how natural it looks. It would have been nice to see a quantitative comparison to mocap data for the same tasks.

(c) Reproducibility: This paper combines several advance components. While it would be possible to implement the same components, I'm skeptical that a different implementation would result in the exact same results. The results would be similar but the inherent complexity of the system will result in small differences. The paper does document limits and other parameters of the system.

(d) Improvements in Research/Writing: Overal it is a very clear paper. The paper combines many techniques and therefore focus on the integration of the techniques. Therefore it feels a bit like a survey where the the reader has to follow up with the citations. I don't think it is the fault of the paper but it could have flushed out more concepts along the way (for example the jacobian transpose method). However this might have made the paper too long and even harder to read.

-- DavidMatheson - 24 Nov 2011

Contribution: The paper presenets a new method for biped walking control that is generalizable across different skills, gait parameters, and styles of motion. They choose four control components, which as shown by the authors, reinforce each other leading to a more generalizable control strategy. The control strategy is real-time for a single character.

Evaluation: The approach was tested on many different motions, motion styles and characters, including carrying heavy object, going up the stairs, crowd simulation, etc. Overall it looks that the evaluation was pretty extensive.

Reproducibility: It looks like all the details of the implementation are there, but I wouldn't really try to implement it: looks quite complicated.

Improvements: It would be interesting to see how well the model generalizes for different terrains and/or physical conditions (environmental effects: wind/water/...) - it looks like it can be modified to be generalized to that.

-- MikhailBessmeltsev - 24 Nov 2011

The paper introduced a multi-component biped walking controller that can be generalized for a variety of high-level tasks. The interaction of various components, such as the feedback and compensation components, allow the same controller to refine and adapt itself to work for a different body shapes and changing movement trajectories in real-time that would not have been possible for each component alone.

The results were generally qualitative evaluations of the controller adapted to perform a variety of tasks and skills without the much-dreaded parameter tuning normally required of other controllers. A general robustness test was also done by pushing walking bodies of various types. A usability test was done for an interactive version of this system allowing for realtime style and user-controlled parameter modification even by a novice.

The work appears reproducible, in the sense that the full working source code is available online. Perhaps this explains why the researchers spent little time explaining implementation details and more about the motivation and conceptual involved, which can appear rather abstract to the unfamiliar reader. A more grounded perspective in an appendix would have been helpful.

-- KevinWoo

Contribution :

parameterized biped walking that allows realtime control is suggested. Character styles (i.e. shape and structure) can be changed in realtime. Generalized control for variety of walking based skills are also developed.

Results Evaluation:

The basic balance control parameters are determined based on the control framework. Other factors are determined for optimum performace viz., friction coefficient, character mass, PD control gain etc. A human model containing a total DOF of 37 (including 6 for global positioning) is used. Dynamics are simulated using ODE with a time step of .0005 seconds. Results corresponding to generalization accross gait, style, characters and tasks are computed and explained.

Reproducibility:

I believe the paper is reproducible. All the parameter selections are backed with proper explainations. Although, exact reproduction should be time consuming given the task of learning the techniques that have been referred to other papers, for instance the invert pendulum model (IPM) and SIMBICON.

Improvements:

Although, overall the writing style is good and up to point. Some of the text could have been swapped out by pseudo code and/or algorithmic description of methodology.

-- Main.sumanm - 24 Nov 2011

Contribution The paper presents an easily-generalizable control scheme for physically-based walking bipeds. The control scheme incorporates as components a set of previously existing control strategies that are well-developed and understood. Furthermore, a system for non-expert authoring of walking motions is demonstrated.

Evaluation The system is evaluated by demonstrating the generalization of the control scheme over various different aspects of a virtual walking character: the gait, the walking style, the character body shape. The control scheme is also evaluated using a number of task-based simulations such as reaching for an object, navigating uneven terrain and obstacles, and even moving a heavy object.

Reproducibility The paper is very much reproducible due to both the detailed description of the algorithms involved, as well as the freely available source code.

Improvements The paper is well-organized and well-written, making it easy to understand the concepts presented within. I have no significant criticisms of the research content either.

-- Main.cdoran - 24 Nov 2011

Contribution
This study combines four components in a novel way that allows the proposed model to achieve a control mechanism for physically-simulated walking motions. As a result, the controller generalizes across gait parameters, motion styles, character proportions, and a variety of skills. It also provides a user-friendly tool such that even novice users can create a character of desired proportions and see the resulting motion for that character immediately.

Evaluation
The results are evaluated with different gait parameters, i.e. walking forwards-backwards, varying walking speeds, and stepping frequencies, and turning towards a desired direction. The study shows the control over gait parameters for characters with different body types and for different styles. Moreover, generalization of several motion styles are also shown with the accompanying videos. It is alsoshown that the framework is capable of generalizing across a variety of characters with varying proportions and tasks such as reaching, pulling/pushing a crate, etc. The user interface for the authoring of the motion styles and character proportions is also tested by novice users for its ease of usability. The framework is tested under different parameters, such as with doubled and halved PD-gains, varying alpha values for the IPM. Lastly, contributions of each control component is analyzed and shown in the provided videos.

Reproducibility
The framework is reproducible, as the source code is open-source, and the methodology of the implementation is detailed enough to follow.

Improvements
In general, the paper is well-organized. The sections are divided meaningfully, and easy to follow.

ozgur
-- Main.ooguz - 24 Nov 2011

(a) What is the contribution of the paper? The authors present a generalized real-time control strategy for walking bipedal characters. Specific tuning with respect to parameters or characters is not required, the controller is robust to disturbances, and control strategies can be successfully authored by non-expert users. The research integrates a series of constituent components that have existed in the control literature for 16 years. (b) How are the results evaluated? Results are provided in the form of two videos. Results show that the proposed walking control models were able to generalize across (1) gait parameters (both forward and backward), (2) styles (authored by novice users), (3) characters (with arbitrary proportions) and (4) tasks (such as reaching, moving a crate, navigating around obstacles/stairs and in crowds). (c) Is the paper reproducible? Yes. Source code is provided by the authors, along with all implementation details (parameters, etc.) required to reproduce the simulation in ODE. (d) How could the paper research or paper writing be improved? No complaints here.

-- DanielTroniak - 24 Nov 2011

Generalized Biped Walking Control

Contribution:
Evaluation:
Reproducibility:
Improvements:

-- MichielVanDePanne - 24 Nov 2011

<--/commentPlugin-->