Conventional ergonomics analysis involves building physical mockups and conducting simulated operations, such that the constraints experienced by the human subjects can be directly observed. The limitations of this approach are that, they are resource intensive, less flexible for testing design variability and difficult to involve large number of subjects to account for population variability and thus, it is a reactive approach. With the advent of computer aided techniques, efforts are on to support ergonomics analysis processes for proactive design approaches. To achieve this, real scenarios are being simulated in virtual environments which include induction of representative human subjects into such envi-ronments and are termed as Digital Human Models (DHMs). The main challenge in the simulation of humans is to obtain the naturalness that is perceived in human interaction with the environment. This naturalness can be achieved by synergetically modeling the physical performance and cognitive aspects of humans in such a way that one aspect caters the requirements posed from the other. But in current DHMs, the various elements in the physical performance aspect are not in line with the requirements of higher level behav¬ioral/cogntive aspects. Towards meeting this objective, the influence of physical performance aspects of humans on achieving naturalness when DHM interacts with the virtual environment has been studied. In this work, the task of ’reach’ has been chosen for studying the influence of kinematic structure, posture modeling and stability aspects on achieving naturalness for both discrete and statistical humans. Also, a framework has been developed to give instructions based on relations between the segments of the body and objects in the environment. Kinematic structure is modeled to simulate the humans with varied dimensions taking care of the change of link fixations necessary for various tasks. The conventional techniques used to define kinematic structures have limitations in resolving the issues that arise due to change in link fixations. In this work a new scheme is developed to effectively handle precedence relationship sand change of configuration of the existing posture whenever link fixations change. The advantage with this new approach is that complex maneuvers which involve different link fixations and multiple fixations at a time can be managed automati¬cally without the user’s intervention. Posture prediction involves estimation of the whole body posture which a human operator is likely to assume while performing a task. It involves finding a configuration satisfy¬ing the constraints like placing the body-segments in preferred locations of the task space and satisfying the relations specified between body segments. There are two main chal¬lenges in this regard; one is achieving naturalness in the predicted postures and the other is minimizing the mathematical complexity involved in finding the real time solutions. A human-specific posture prediction framework is developed which can handle a variety of constraints and realize the natural behavior. The approach is completely geometry based and unlike numerical methods, the solutions involve no matrix inversions. Digital human models (DHMs), both as avatars and agents, need to be controlled to make them manipulate the objects in the virtual world. A relations based description scheme is developed to instruct the DHM to perform the tasks. The descriptions as a set of relations and postures involve simple triplets and quadruplets. As the descriptions constitute only the relations between actors, incorporating different behavior models while executing the relations is feasible through this framework. Static balancing is one of the crucial factors influencing the posture of humans. The stim¬ulus for the static balancing is the body’s self weight and is governed by the location of its point of application, namely the center of mass (COM). The main focus is on determin¬ing suitable locations for COM to infer about the mobility of the segments which supports the human structure in slow motion scenarios. Various geometric conditions necessary for support retaining, altering are deduced and developed strategies for posture transitions for effective task performance while maintaining stability. These conditions are useful in de¬termining the posture transition required to shift the COM from one region to the other and thus the behaviors realized while accomplishing the tasks are realistic. These behaviors are simulated through statically stable walking and sit to stand posture transition. One of the advantages of employing DHMs in virtual simulations is the feasibility of creat¬ing human models with varied dimensions. A comparative study is conducted on different methods based on probabilistic and statistic theory as an alternative to the percentile based approach with a view to answer the questions like ’what percentage of people can success-fully accomplish a certain task’ and ’how well can people perform when they reach a point in the operational space’. The case study is done assuming upper and lower arms of hu¬mans as a two link planar manipulator and their link lengths as random variables. Making use of statistical DHMs, the concept of task dependent boundary manikins is introduced to geometrically characterize the extreme individuals in the given population who would ac-complish the task. Simulations with these manikins would help designers to visualize how differently the extreme individuals would perform the task. All these different aspects of DHM discussed are incorporated in our native DHM developed named ’MAYAMANAV’. Finally this thesis will end with conclusions and future work discussing how these different aspects of DHM discussed can be combined with behavioral models to simulate the human error.