Munjal Desai

Impact of robot failures and feedback on real-time trust

Prior work in human trust of autonomous robots suggests the timing of reliability drops impact trust and control allocation strategies. However, trust is traditionally measured post-run, thereby masking the real-time changes in trust, reducing sensitivity to factors like inertia, and subjecting the measure to biases like the primacy-recency effect. Likewise, little is known on how feedback of robot confidence interacts in real-time with trust and control allocation strategies. An experiment to examine these issues showed trust loss due to early reliability drops is masked in traditional post-run measures, trust demonstrates inertia, and feedback alters allocation strategies independent of trust. The implications of specific findings on development of trust models and robot design are also discussed.

Essential features of telepresence robots

Telepresence robots are mobile robot platforms capable of providing two way audio and video communication. Recently there has been a surge in companies designing telepresence robots. We conducted a series of user studies at Google in Mountain View with two different commercially available telepresence robots. Based on the data collected from these user studies, we present a set of guidelines for designing telepresence robots. These essential guidelines pertain to video, audio, user interface, physical features, and autonomous behaviors.

Analysis of natural gestures for controlling robot teams on multi-touch tabletop surfaces

Multi-touch technologies hold much promise for the command and control of mobile robot teams. To improve the ease of learning and usability of these interfaces, we conducted an experiment to determine the gestures that people would naturally use, rather than the gestures they would be instructed to use in a pre-designed system. A set of 26 tasks with differing control needs were presented sequentially on a DiamondTouch to 31 participants. We found that the task of controlling robots exposed unique gesture sets and considerations not previously observed, particularly in desktop-like applications. In this paper, we present the details of these findings, a taxonomy of the gesture set, and guidelines for designing gesture sets for robot control.

Blending human and robot inputs for sliding scale autonomy

Most robot systems have discrete autonomy levels, if they possess more than a single autonomy level. A user or the robot may switch between these discrete modes, but the robot can not operate at a level between any two modes. We have developed a sliding scale autonomy system that allows autonomy levels to be created and changed on the fly. This paper discusses the systems architecture and presents the results of experiments with the sliding scale autonomy system.

Robot confidence and trust alignment

Trust in automation plays a crucial role in human-robot interaction and usually varies during interactions. In scenarios of shared control, the ideal pattern is for the users real-time trust in the robot to align with robot performance. This should lead to an increased overall efficiency of the system by limiting under-trust and over-trust. However, users sometimes display incorrect trust and the ability to detect and alter user trust is important. This paper describes measures for real-time trust alignment.

Potential measures for detecting trust changes

It is challenging to quantitatively measure a users trust in a robot system using traditional survey methods due to their invasiveness and tendency to disrupt the flow of operation. Therefore, we analyzed data from an existing experiment to identify measures which (1) have face validity for measuring trust and (2) align with the collected post-run trust measures. Two measures are promising as real-time indications of a drop in trust. The first is the time between the most recent warning and when the participant reduces the robots autonomy level. The second is the number of warnings prior to the reduction of the autonomy level.

Considering the bystander's perspective for indirect human-robot interaction

As robots become more pervasive in society, people will find themselves actively interacting with robots, and also rushing past them without any explicit interaction. People are able to maneuver in crowded situations by speeding up or slowing down to slip in between open pockets where people are not standing or walking. Our research focuses on this indirect bystander interaction. Scholtz defines a bystander as a person who “does not explicitly interact with a robot but needs some model of robot behavior to understand the consequences of the robots actions” and does not have formal training about the robot [1], [2]. We investigated the level of trust that a bystander has of a robotic system in a corridor passing scenario by asking people to watch short videos of such scenarios where the hallway is only wide enough to accommodate two entities (either human or robot). Our goal was to understand the bystanders mental model of how a robot should behave when passing a human, the bystanders expectation of the robot to adhere to social protocol, and the overall trust a bystander has of the robot to do the right thing.

Towards measuring the quality of interaction: communication through telepresence robots

Personal video conferencing is now a common occurrence in long distance interpersonal relationships. Telepresence robots additionally provide mobility to video conferencing, and people can converse without being restricted to a single vantage point. The metrics to explicitly quantify person to person interaction through a telepresence robot do not yet exist. In this paper, we discuss technical requirements needed to support such a communication. We also look at the fields of human-computer interaction (HCI), computer supported cooperative work (CSCW), communications, and psychology for quantitative and qualitative performance measures which are independent of interpersonal relationships and communication task.

Hand and finger registration for multi-touch joysticks on software-based operator control units

Robot control typically requires many physical joysticks, buttons, and switches. Taking inspiration from video game controllers, we have created a Dynamically Resizing, Ergonomic, and Multi-touch (DREAM) controller to allow for the development of a software-based operator control unit (SoftOCU). The DREAM Controller is created wherever a person places his or her hand; thus we needed to develop an algorithm for accurate hand and finger registration. Tested with a set of 405 hands from 62 users, our algorithm correctly identified 97% of the hands.

Using the "negative attitude toward robots scale" with telepresence robots

Nomura et al. designed and piloted the Negative Attitude toward Robots Scale (NARS) in 2003 [18]. NARS has been used by researchers to understand the attitudes of different people towards robots under different circumstances. To our knowledge, NARS has only been used with robots perceived to be autonomous. Our goal was to evaluate if NARS could be extended to robots that were known to have a human in the loop. Towards this end, we verified the validity of the scale with telepresence robots using an online video survey. We found differences across different cultures and gender much like other researcher in the past. Once the consistency of the scale was verified, we used NARS in our second study that involved the use of telepresence robots.